Enter Solaris 8 brandz zones. Being able to run legacy applications in a Solaris 8 brandz generally gets you around most of the issues in the remaining apps that do stupid things. However, there’s a couple of quirks with a zone that can introduce odd behaviour in an application.

Progress for example.…… Read On.

The issue

An old legacy Progress database was sitting on an old legacy server which was so EOL that it would make even Bill cringe… or laugh… Running Solaris 6 was not an option, but running anything else, (even Solaris 9), would be adventurous to say the least. Can I say legacy again?

The solution

So first try a brandz Solaris 8. Configured, installed, and booted a brand spanking new zone. Copy the data over. Attempt a start of the Progress database.

10:03:21 BROKER  0: Multi-user session begin. (333)
10:03:24 BROKER  0: Begin Physical Redo Phase at 13312 . (5326)
10:03:30 BROKER  0: Physical Redo Phase Completed at blk 13743 off 2457 upd 2636. (7161)
10:03:31 BROKER  0: Started for mfglive using TCP, pid 20533. (5644)
10:04:02 BROKER  0: SYSTEM ERROR: Unable to kill parent process, errno= 4. (1680)
10:04:02 BROKER  0: SYSTEM ERROR: The broker is exiting unexpectedly, beginning Abnormal Shutdown. (5292)
10:04:02 BROKER  0: drexit: Initiating Abnormal Shutdown
10:04:02 BROKER  0: ** Save file named core for analysis by Progress Software Corporation. (439)
10:04:02 BROKER  0: Begin ABNORMAL shutdown code 2 (2249)
10:04:04 BROKER   : Removed shared memory with segment_id: 41
10:04:04 BROKER   : Multi-user session end. (334)

Nada. Mmmmm. Seems to start up OK, but then crashes in a screaming heap a short while later.

Checking shared libraries seemed to be OK. Nothing missing or odd.

root % ldd -r /mfgpro/dlc91c/bin/_mprosrv
        /usr/lib/secure/s8_preload.so.1
        libsocket.so.1 =>        /usr/lib/libsocket.so.1
        libnsl.so.1 =>   /usr/lib/libnsl.so.1
        libintl.so.1 =>  /usr/lib/libintl.so.1
        libdl.so.1 =>    /usr/lib/libdl.so.1
        libm.so.1 =>     /usr/lib/libm.so.1
        libthread.so.1 =>        /usr/lib/libthread.so.1
        libc.so.1 =>     /usr/lib/libc.so.1
        libmp.so.2 =>    /usr/lib/libmp.so.2
        /usr/platform/SUNW,Sun-Fire-V490/lib/libc_psr.so.1

So, I thought I’d truss, (the old friend), a few processes to see what was going on. Chucking the below in various scripts, and comparing the working server to the non-working server. (This produces VERY NOISY output, but it’s great for diagnosing what’s going on.)

truss -feal -vall -xall -rall -wall -sall -mall -o /tmp/logfile exec_process

If you look at the appendices to this post, I’ve added the relevant, sections of the truss. You may notice an interesting thing. The working server roughly follows this process, (in the snippet):
– Bind to 0.0.0.0 port 3500.
– fstat64() filehandles.
– Yadda, yadda, write output to STDOUT.
– getpid() — find parent process, (24012).
– kill –TERM 24012
– wait()
– Parent — ALRM signal
– Parent — die.
– Child — Context switch.
– getpid() — find parent process, (1). OK!

Now this is a little odd. It would seem that some developer thought that it’d be a damn good idea to re-invent the old “I’m going to fork a child process in the background” API calls, and chuck in their own.

They are checking the parent process, killing it, and rechecking to see if it’s now 1, (which is sysvinit), then it progresses on.

Now, the non-working server.
– Bind to 0.0.0.0 port 3500.
– fstat64() filehandles.
– Yadda, yadda, write output to STDOUT.
– getpid() — find parent process, (21267).
– kill –TERM 21267
– wait()
– Parent — ALRM signal <-IMPORTANT
– Parent — no die.
– getpid() — find parent process, (21267).
– kill –TERM 21267
– wait()
– Parent — ALRM signal <-IMPORTANT
– Parent — no die.
Ad naseum.….

A quick process check:

     UID   PID  PPID  C    STIME TTY      TIME CMD
    root  2260  1417  0   May 27 ?        0:09 /etc/init
    root  1417  1417  0   May 27 ?        0:00 zsched
    root  2960  1417  0   May 27 ?        0:00 /usr/lib/picl/picld

Hey! Of course you’re not gonna be able to kill zsched! You see in a Solaris zone the ‘root’ process is called ‘zsched’, and it will never ever be 1 at all, and neither will init for that matter. Not only that you’re not ever going to be able to kill the zsched process from within the zone as it’s protected within the global zone memory space.

Well, it’s too late to kick the developer up the bum. So time to hack.

The hack

Preloading shared libs has been around since I was a Mars Bar in my Dad’s top pocket, (well almost). So it should be a simple case of replacing any getpid() API call with my own. Simple enough, done it before.

See the appendix for the getpidhack.c file that I used. This is a shared lib that replaces both getppid() and getpid(), (for good measure, but not really needed in my case), API calls. It will return the PID 1, when _getppid() returns the same PID as the environment variable PIDHACK.
Note that I call _getppid() which allows me to fetch the REAL PID so I can return it when they don’t match.

So, to use it. First build it:

gcc -fPIC -c getpidhack.c
gld getpidhack.o -L/usr/lib -lc -G -assert pure-text -o /usr/local/lib/libgetpidhack.so

Set an environment variable called PIDHACK to the currently running zsched process:

HACKPID=`ps -e | awk '/zsched/{print$1}'`
export HACKPID

Then set your LD_PRELOAD variable to point to the shared lib:

LD_PRELOAD="/usr/local/lib/libgetpidhack.so"
export LD_PRELOAD

After this the Progress database just started. Woot! The DB can now run on hardware that it would never in it’s life be able to run on under a brandz zone. Another EOL server to chuck out the computer room window to see what happens.

BTW, if you are having issues with this, you can run a DTrace to see if you’re calling the real getpid(), or the getpidhack version. Create the following script, and run it with one argument, (the PID of your zsched). Every time you hit the getppid() API call it’ll register — this means that you are not replacing the API call and your shared lib preload isn’t working.

#!/usr/sbin/dtrace -s

syscall::getpid:return
/ppid == $1/
{
        printf("%d %d", ppid, pid);
}

Appendix 1 — getpidhack.c

#include <sys/types.h>
#include <unistd.h>
#include <stdio.h>

pid_t getppid(void)
{
        pid_t hackppid;
        pid_t realppid;

        // hackpid - the PID of the zsched process.
        hackppid = atol(getenv("HACKPID"));

        // realppid - the real PID of the parent process.
        realppid = _getppid();

        // If the realppid equals hackppid, then we want to reset it to 1.
        if (realppid == hackppid)
        {
                return((pid_t) 1);
        }

        // Else return whatever PID we were given.
        return(realppid);
}

pid_t getpid(void)
{
        pid_t hackpid;
        pid_t realpid;

        // hackpid - the PID of the zsched process.
        hackpid = atol(getenv("HACKPID"));

        // realpid - the real PID of the parent process.
        realpid = _getpid();

        // If the realpid equals hackpid, then we want to reset it to 1.
        if (realpid == hackpid)
        {
                return((pid_t) 1);
        }

        // Else return whatever PID we were given.
        return(realpid);
}

Appendix 2 — The working server

24013/1:        setsockopt(76, 65535, 4, 0xEFFFF7FC, 4)         = 0
24013/1:        setsockopt(76, 65535, 8, 0xEFFFF7FC, 4)         = 0
24013/1:        bind(76, 0x0015DCA8, 16)                        = 0
24013/1:                name = 0.0.0.0/3500
24013/1:        listen(76, 10)                                  = 0
24013/1:        getpid()                                        = 24013 [24012]
24013/1:        lseek(3, 457164, 0)                             = 457164
24013/1:        read(3, 0xEFFFF66C, 81)                         = 81
24013/1:           % L S t a r t e d   f o r   % s   u s i n g   % s ,   p i d  %
24013/1:           l .   ( 5 6 4 4 )\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0 \0
24013/1:          \0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0
24013/1:        time()                                          = 1252542477
24013/1:        write(9, 0xEFFFF154, 69)                        = 69
24013/1:           1 0 : 2 7 : 5 7   B R O K E R     0 :   S t a r t e d   f o r
24013/1:           m f g l i v e   u s i n g   T C P ,   p i d   2 4 0 1 3 .   ( 5
24013/1:           6 4 4 )\n
24013/1:        fstat64(0, 0xEFFFF058)                          = 0
24013/1:            d=0x01540000 i=17139 m=0020620 l=1  u=0     g=7     rdev=0x0 0600005
24013/1:                at = Sep 10 10:27:46 EST 2009  [ 1252542466 ]
24013/1:                mt = Sep 10 10:27:56 EST 2009  [ 1252542476 ]
24013/1:                ct = Sep 10 09:39:31 EST 2009  [ 1252539571 ]
24013/1:            bsz=8192  blks=0     fs=ufs
24013/1:        fstat64(1, 0xEFFFF058)                          = 0
24013/1:            d=0x01540000 i=17139 m=0020620 l=1  u=0     g=7     rdev=0x0 0600005
24013/1:                at = Sep 10 10:27:46 EST 2009  [ 1252542466 ]
24013/1:                mt = Sep 10 10:27:56 EST 2009  [ 1252542476 ]
24013/1:                ct = Sep 10 09:39:31 EST 2009  [ 1252539571 ]
24013/1:            bsz=8192  blks=0     fs=ufs
24013/1:        fstat64(2, 0xEFFFF058)                          = 0
24013/1:            d=0x01540000 i=17139 m=0020620 l=1  u=0     g=7     rdev=0x0 0600005
24013/1:                at = Sep 10 10:27:46 EST 2009  [ 1252542466 ]
24013/1:                mt = Sep 10 10:27:56 EST 2009  [ 1252542476 ]
24013/1:                ct = Sep 10 09:39:31 EST 2009  [ 1252539571 ]
24013/1:            bsz=8192  blks=0     fs=ufs
24013/1:        write(1, 0xEFFFF154, 69)                        = 69
24013/1:           1 0 : 2 7 : 5 7   B R O K E R     0 :   S t a r t e d   f o r
24013/1:           m f g l i v e   u s i n g   T C P ,   p i d   2 4 0 1 3 .   ( 5
24013/1:           6 4 4 )\n
24013/1:        pwrite64(7, 0xEFFFF908, 4, 0)                   = 4
24013/1:             0xEFFFF908: "\0\0\002"
24013/1:        close(7)                                        = 0
24013/1:        getpid()                                        = 24013 [24012]
24013/1:        getpid()                                        = 24013 [24012]
24013/1:        kill(24012, 0x0000000F)                         = 0
24012/2:        signotifywait()                                 = 15
24013/1:        lwp_alarm(1)                                    = 0
24012/2:            Incurred fault #11, FLTPAGE  %pc = 0xEF5CDC80  addr = 0xEF67 0F48
24012/2:            Incurred fault #11, FLTPAGE  %pc = 0xEF64A3C8  addr = 0xEF66 FA68
24012/2:        lwp_sigredirect(1, 0x0000000F)                  = 0
24012/1:            Received signal #15, SIGTERM, in wait() [caught]
24012/1:              siginfo: SIGTERM pid=24013 uid=0
24012/1:        wait()                                          Err#4 EINTR
24012/1:        sigprocmask(3, 0xEF667DF8, 0x00000000)          = 0
24012/1:                 set = 0 0 0 0
24012/1:            Incurred fault #11, FLTPAGE  %pc = 0x0004593C  addr = 0x0013 10F4
24012/1:            Incurred fault #11, FLTPAGE  %pc = 0x000459A8  addr = 0x0015 3AEC
24012/1:        _exit(0)
24013/1:            Received signal #14, SIGALRM, in pause() [caught]
24013/1:        pause()                                         Err#4 EINTR
24013/1:            Incurred fault #11, FLTPAGE  %pc = 0xEF5CDCE8  addr = 0xEF5C DCE8
24013/1:        sigprocmask(3, 0xEF667DF8, 0x00000000)          = 0
24013/1:                 set = 0 0 0 0
24013/1:        context(1, 0xEFFFF240)
24013/1:        getpid()                                        = 24013 [1]
24013/1:        close(0)                                        = 0
24013/1:        open(0xEFFFF920, 0)                             = 0
24013/1:             0xEFFFF920: "/qaddb/eblive/mfglive.lg"
24013/1:        close(1)                                        = 0
24013/1:        open(0xEFFFF920, 0)                             = 1
24013/1:             0xEFFFF920: "/qaddb/eblive/mfglive.lg"
24013/1:        close(2)                                        = 0
24013/1:        open(0xEFFFF920, 0)                             = 2
24013/1:             0xEFFFF920: "/qaddb/eblive/mfglive.lg"
24013/1:        lseek(3, 342954, 0)                             = 342954
24013/1:        read(3, 0xEFFFF570, 81)                         = 81
24013/1:           % L P R O G R E S S   V e r s i o n   % s   o n   % s .   ( 4 2
24013/1:           3 4 )\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0 \0
24013/1:          \0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0
24013/1:        time()                                          = 1252542478
24013/1:        write(9, 0xEFFFF05C, 61)                        = 61
24013/1:           1 0 : 2 7 : 5 8   B R O K E R     0 :   P R O G R E S S   V e r
24013/1:           s i o n   9 . 1 C   o n   S O L A R I S .   ( 4 2 3 4 )\n
24013/1:        fstat64(0, 0xEFFFEF60)                          = 0
24013/1:            d=0x01540002 i=68332 m=0100644 l=1  u=0     g=1     sz=97192 88
24013/1:                at = Sep 10 03:01:28 EST 2009  [ 1252515688 ]
24013/1:                mt = Sep 10 10:27:58 EST 2009  [ 1252542478 ]
24013/1:                ct = Sep 10 10:27:58 EST 2009  [ 1252542478 ]
24013/1:            bsz=8192  blks=19008 fs=ufs
24013/1:        fstat64(1, 0xEFFFEF60)                          = 0
24013/1:            d=0x01540002 i=68332 m=0100644 l=1  u=0     g=1     sz=97192 88
24013/1:                at = Sep 10 03:01:28 EST 2009  [ 1252515688 ]
24013/1:                mt = Sep 10 10:27:58 EST 2009  [ 1252542478 ]
24013/1:                ct = Sep 10 10:27:58 EST 2009  [ 1252542478 ]
24013/1:            bsz=8192  blks=19008 fs=ufs
24013/1:        fstat64(2, 0xEFFFEF60)                          = 0
24013/1:            d=0x01540002 i=68332 m=0100644 l=1  u=0     g=1     sz=97192 88
24013/1:                at = Sep 10 03:01:28 EST 2009  [ 1252515688 ]
24013/1:                mt = Sep 10 10:27:58 EST 2009  [ 1252542478 ]
24013/1:                ct = Sep 10 10:27:58 EST 2009  [ 1252542478 ]
24013/1:            bsz=8192  blks=19008 fs=ufs
24013/1:        write(1, 0xEFFFF05C, 61)                        Err#9 EBADF
24013/1:           1 0 : 2 7 : 5 8   B R O K E R     0 :   P R O G R E S S   V e r
24013/1:           s i o n   9 . 1 C   o n   S O L A R I S .   ( 4 2 3 4 )\n
24013/1:        lseek(3, 346761, 0)                             = 346761
24013/1:        read(3, 0xEFFFF570, 81)                         = 81
24013/1:           % L S e r v e r   s t a r t e d   b y   % s   o n   % s .   ( 4
24013/1:           2 8 1 )\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0 \0
24013/1:          \0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0
24013/1:        time()                                          = 1252542478
24013/1:        write(9, 0xEFFFF05C, 65)                        = 65
24013/1:           1 0 : 2 7 : 5 8   B R O K E R     0 :   S e r v e r   s t a r t
24013/1:           e d   b y   r o o t   o n   / d e v / p t s / 5 .   ( 4 2 8 1)
24013/1:          \n

Appendix 3 — The non-working server

21268/1:        setsockopt(76, 65535, 4, 0xFFBFF9AC, 4, 1)      = 0
21268/1:        setsockopt(76, 65535, 8, 0xFFBFF9AC, 4, 1)      = 0
21268/1:        bind(76, 0x0015DCA8, 16, 3)                     = 0
21268/1:                AF_INET  name = 0.0.0.0  port = 3500
21268/1:        listen(76, 10, 1)                               = 0
21268/1:        getpid()                                        = 21268 [21267]
21268/1:        lseek(3, 457164, 0)                             = 457164
21268/1:        read(3, 0xFFBFF81C, 81)                         = 81
21268/1:           % L S t a r t e d   f o r   % s   u s i n g   % s ,   p i d   %
21268/1:           l .   ( 5 6 4 4 )\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0
21268/1:          \0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0
21268/1:        time()                                          = 1252542489
21268/1:        write(9, 0xFFBFF304, 69)                        = 69
21268/1:           1 0 : 2 8 : 0 9   B R O K E R     0 :   S t a r t e d   f o r
21268/1:           m f g l i v e   u s i n g   T C P ,   p i d   2 1 2 6 8 .   ( 5
21268/1:           6 4 4 )\n
21268/1:        fstat64(0, 0xFFBFF208)                          = 0
21268/1:            d=0x05980009 i=4426329 m=0020620 l=1  u=0     g=7     rdev=0x00600011
21268/1:                at = Sep 10 10:27:57 EST 2009  [ 1252542477 ]
21268/1:                mt = Sep 10 10:28:09 EST 2009  [ 1252542489 ]
21268/1:                ct = Sep 10 09:44:02 EST 2009  [ 1252539842 ]
21268/1:            bsz=8192  blks=0     fs=lofs
21268/1:        fstat64(1, 0xFFBFF208)                          = 0
21268/1:            d=0x05980009 i=4426329 m=0020620 l=1  u=0     g=7     rdev=0x00600011
21268/1:                at = Sep 10 10:27:57 EST 2009  [ 1252542477 ]
21268/1:                mt = Sep 10 10:28:09 EST 2009  [ 1252542489 ]
21268/1:                ct = Sep 10 09:44:02 EST 2009  [ 1252539842 ]
21268/1:            bsz=8192  blks=0     fs=lofs
21268/1:        fstat64(2, 0xFFBFF208)                          = 0
21268/1:            d=0x05980009 i=4426329 m=0020620 l=1  u=0     g=7     rdev=0x00600011
21268/1:                at = Sep 10 10:27:57 EST 2009  [ 1252542477 ]
21268/1:                mt = Sep 10 10:28:09 EST 2009  [ 1252542489 ]
21268/1:                ct = Sep 10 09:44:02 EST 2009  [ 1252539842 ]
21268/1:            bsz=8192  blks=0     fs=lofs
21268/1:        write(1, 0xFFBFF304, 69)                        = 69
21268/1:           1 0 : 2 8 : 0 9   B R O K E R     0 :   S t a r t e d   f o r
21268/1:           m f g l i v e   u s i n g   T C P ,   p i d   2 1 2 6 8 .   ( 5
21268/1:           6 4 4 )\n
21268/1:        pwrite64(7, 0xFFBFFAB8, 4, 0)                   = 4
21268/1:             0xFFBFFAB8: "\0\0\002"
21268/1:        close(7)                                        = 0
21268/1:        getpid()                                        = 21268 [21267]
21268/1:        kill(21267, 0x0000000F)                         = 0
21267/1:            Received signal #15, SIGTERM, in wait() [caught]
21267/1:              siginfo: SIG#0
21268/1:            Incurred fault #11, FLTPAGE  %pc = 0x000C3048  addr = 0x58900137234
21267/1:        wait()                                          Err#4 EINTR
21268/1:        alarm(1)                                        = 0
21267/1:        lwp_sigtimedwait(0xFFBFF570, 0xFFBFF498, 0x00000010) = 0
21267/1:                sigmask = 0 0 0 0
21267/1:              siginfo: SIG#0
21267/1:        lwp_sigtimedwait(0xFFBFF488, 0xFFBFF570, 0x00000010) = 0
21267/1:                sigmask = 0 0 0 0
21267/1:              siginfo: SIG#0
21267/1:        lwp_sigtimedwait(0xFFBFF484, 0xFFBFF2B8, 0x00000010) = 0
21267/1:                sigmask = 0x00004000 0 0 0
21267/1:              siginfo: SIG#16384
21267/1:        lwp_sigtimedwait(0xFFBFF2A8, 0xFFBFF330, 0x00000010) = 0
21267/1:                sigmask = 0x00004000 0 0 0
21267/1:              siginfo: SIG#16384
21267/1:        sigprocmask(3, 0xFFBFF330, 0x00000000)          = 0
21267/1:                 set = 0x00004000 0 0 0
21267/1:            Incurred fault #11, FLTPAGE  %pc = 0x0004593C  addr = 0x589001310F4
21267/1:            Incurred fault #11, FLTPAGE  %pc = 0x000459A8  addr = 0x58900153AEC
21267/1:        _exit(0)
21268/1:            Received signal #14, SIGALRM, in pause() [caught]
21268/1:              siginfo: SIG#0
21268/1:        pause()                                         Err#4 EINTR
21268/1:        lwp_sigtimedwait(0xFFBFF698, 0xFFBFF5C0, 0x00000010) = 0
21268/1:                sigmask = 0 0 0 0
21268/1:              siginfo: SIG#0
21268/1:        lwp_sigtimedwait(0xFFBFF5B0, 0xFFBFF698, 0x00000010) = 0
21268/1:                sigmask = 0 0 0 0
21268/1:              siginfo: SIG#0
21268/1:        lwp_sigtimedwait(0xFFBFF5AC, 0xFFBFF3E0, 0x00000010) = 0
21268/1:                sigmask = 0x00002000 0 0 0
21268/1:              siginfo: SIG#8192
21268/1:        lwp_sigtimedwait(0xFFBFF3D0, 0xFFBFF458, 0x00000010) = 0
21268/1:                sigmask = 0x00002000 0 0 0
21268/1:              siginfo: SIG#8192
21268/1:        sigprocmask(3, 0xFFBFF458, 0x00000000)          = 0
21268/1:                 set = 0x00002000 0 0 0
21268/1:        lwp_sigtimedwait(0xFF1D7C18, 0xFFBFF1B0, 0x00000010) = 0
21268/1:                sigmask = 0xFFBFFEFF 0x00001FFF 0 0
21268/1:              siginfo: SIG#-4194561
21268/1:        lwp_sigtimedwait(0xFFBFF1A0, 0xFFBFF228, 0x00000010) = 0
21268/1:                sigmask = 0xFFBFFEFF 0x0000FF1F 0 0
21268/1:              siginfo: SIG#-4194561
21268/1:        sigprocmask(3, 0xFFBFF228, 0xFFBFF238)          = 0
21268/1:                 set = 0xFFBFFEFF 0x0000FF1F 0 0
21268/1:                oset = 0x00002000 0 0 0
21268/1:        lwp_sigtimedwait(0xFFBFF238, 0xFFBFF1B0, 0x00000010) = 0
21268/1:                sigmask = 0x00002000 0 0 0
21268/1:              siginfo: SIG#8192
21268/1:        lwp_sigtimedwait(0xFFBFF1A0, 0xFFBFF360, 0x00000010) = 0
21268/1:                sigmask = 0x00002000 0 0 0
21268/1:              siginfo: SIG#8192
21268/1:        lwp_park(1, 1, 1)                               = 0
21268/1:        lwp_sigtimedwait(0xFFBFF370, 0xFFBFF088, 0x000001C0) = 0
21268/1:                sigmask = 0x0000082F 0 0 0
21268/1:              siginfo: SIG#2095
21268/1:        lwp_sigtimedwait(0xFFBFF090, 0xFFBFEFF8, 0x00000010) = 0
21268/1:                sigmask = 0 0 0 0
21268/1:              siginfo: SIG#0
21268/1:        lwp_sigtimedwait(0xFFBFEFE8, 0xFFBFF090, 0x00000010) = 0
21268/1:                sigmask = 0 0 0 0
21268/1:              siginfo: SIG#0
21268/1:        context(1, 0xFFBFF088)
21268/1:        getpid()                                        = 21268 [1417]
21268/1:        kill(1417, 0x0000000F)                          = 0
21268/1:        alarm(1)                                        = 0
21268/1:            Received signal #14, SIGALRM, in pause() [caught]
21268/1:              siginfo: SIG#0
21268/1:        pause()                                         Err#4 EINTR

No related posts.

What Prowl does is send “push” events to your iPhone from virtually anything.
Install the app, register on the Prowl website, generate an API key, and you’re home and hosed for some good hacking.

This means that any application that wants to push a notification to your iPhone contacts the Prowl servers, which then sends the notification on to your iPhone. So far it’s reliable enough, but even they say not to trust it for critical or emergency applications.

There is an API, and it is fairly complete, supporting all your language preferences, (Java, perl, PHP, etc, etc — sorry no COBOL).

Of course, there’s even a plugin for WordPress, (which I have to say is so much easier than Joomla). It will send push alerts for new posts, comments, and other things.

WordPress post alert

Looks like someone has posted another blog entry.

Viewing the notification brings you here:

Alerts

Alerts can be adjusted, the nice feature is being able to silence them for a given time.

And adjust what sounds are played and when, although the default sounds are too quiet I find.

Redirects

Redirects allow a specified application to start when a particular event comes in. Trouble is that there’s no way of selecting anything else but the given set of applications. Maybe in the next release.

No related posts.

This was for several reasons:
1. Joomla is a pain in the bum for blogging. Its too complex for a blog and I keep forgetting the differences between a “com”, “module” and a “plugin”.
2. Upgrading is a pain in the bum. I just recently noticed that half my content went missing as part of a recent upgrade.
3. I had bought an iPhone and wanted to be able to post while out and about. Getting this going in Joomla was a pain in the bum, (although I discovered after switching to WordPress that it wasn’t really Joomla’s fault).
4. My old TreeCam blog was running on an old version of s9y, and I wanted to combine it all.

Not that Joomla is all bad, it’s a great CMS app, but just not for blogging. It has a fantastic plugin, (module or whatever the heck you call it), for eCommerce, (virtuemart), and even a blog thingy. I always found it to be a struggle. One of those things. It’s not designed for blogging.

It was a good decision to move. WordPress is infinitely easier than Joomla. None of the learning curve. So far there seems to be the same number of plugins for both, but it looks like Joomla is getting a little “stale”. Installing plugins for WordPress is damn easy. Almost a “one click” install or upgrade for all plugins and templates.

Since the switch I’ve discovered all sorts of goodies. EG out of the box weblog API, RPC calls, blog pings, dead link monitoring.

Nice. Really polished app.

Even better is the fact that I didn’t need to do anything for my iPhone to start posting to my blog. Just install BlogPress, (and the free WordPress app), from the Apple app store, enter in my site, and start blogging. I even used it to migrate from Joomla to WordPress, as you can post to multiple sites in one hit.

BlogPress’ settings page

You can post piccies and videos to Picasa and YouTube. Specify multiple blogs to post to.

Pulling down posts

Posting a story

Is dead easy, although may be a problem for those who don’t know their HTML.

Inserting images

Just click on the camera icon top right, and you can insert photos directly into the post.

Save or post?

You can post it, or save it as a draft for later.

Posting an entry

You’re also given the option of viewing it in Safari, which is a minor annoyance for me. It’d be good to be able to turn it off.

– Posted from my iPhone

No related posts.

Sun’s Java Web Console is a nice little toy. Great for those people who prefer the less extreme learning curve of a CLI, and great for manager types who like pictures.

I’ve used it several times before, but thought I’d try it in on x86 Solaris10u6 box I have here. Bloody Sun. You think a released version implies that it’s stable, and working.

For me the issue was that when I originally installed the server, I used a temporary hostname, and DHCP for network because I couldn’t be bothered to change my local DNS and figure out a spare IP. I then changed to fixed IP and hostname later.

The ramifications of this is that my hostname ended up being assigned to localhost, (aka 127.0.0.1). This means that when the WebConsole was initially installed and configured it used 127.0.0.1 throughout it’s hardcoded, (stupid Tomcat — hate it with a vengence), config files.

This is something that annoys the heck out of me: I understand developers making ‘packages’ that they can ‘deploy’ to a web server, (Tomcat). It’s nice to be able to ‘deploy’ a package. But for heck’s sake, why shove hostname/IP/port definitions in not just one file, but 50 stupid files that are zipped up along with shared libs and binaries? Why? Haven’t they heard of /etc/hosts, or even putting these definitions into somewhere common? Why do they have to have these definitions in multiple places?

Another great example of letting the developer take the lead in systems management.

Anyway, digging around the net trying to resolve my issue showed up a lot of similar complaints, but nothing really to sort it out. ‘wcadmin’ refused to start the ‘deployed’ packages, because they couldn’t connect to the Tomcat server running on my old hostname.

root@somewhere % wcadmin list

Deployed web applications (application name, context name, status):

console  ROOT            [stopped]
console  com_sun_web_ui  [stopped]
console  console         [stopped]
console  manager         [stopped]
console  zfs             [stopped]

Registered jar files (application name, identifier, path):

console  audit_jar     /usr/lib/audit/Audit.jar
console  console_jars  /usr/share/webconsole/lib/*.jar
console  jato_jar      /usr/share/lib/jato/jato.jar
console  javahelp_jar  /usr/jdk/packages/javax.help-2.0/lib/*.jar
console  shared_jars   /usr/share/webconsole/private/container/shared/lib/*.jar

Registered login modules (application name, service name, identifier):

console  ConsoleLogin  userlogin
console  ConsoleLogin  rolelogin

Persistent Jvm options:

-server
-Xmx128m
-XX:+UseParallelGC
-XX:ParallelGCThreads=4

I tried using wcadmin ro ‘undeploy’ and ‘deploy’ packages, and even the brut force method of

root@somewhere % find /var/webconsole /usr/share/webconsole /etc/webconsole /usr/lib/webconsole -type f | xargs egrep -l '127.0.0.1|localhost'

and manually modifying any config files found.

In the end I did a pkgrm/remove configs/pkgadd cycle:

1. Remove WebConsole packages.

system      SUNWasac  Sun Java System Application Server Admin Client
application SUNWmcon  Sun Java(TM) Web Console 3.0.2 (Core)
system      SUNWmconr Sun Java(TM) Web Console 3.0.2 (Root)
application SUNWmcos  Implementation of Sun Java(TM) Web Console (3.0.2) services
application SUNWmcosx Implementation of Sun Java(TM) Web Console (3.0.2) services
application SUNWmctag Sun Java(TM) Web Console 3.0.2 (Tags & Components)
application SUNWzfsgr ZFS Administration for Sun Java(TM) Web Console (Root)
application SUNWzfsgu ZFS Administration for Sun Java(TM) Web Console (usr)

The usual removal:

root@somewhere % pkgrm SUNWzfsgu SUNWmcosx SUNWmcos SUNWasac SUNWmcon SUNWzfsgr SUNWmconr SUNWmctag

Then remove all the config files:

root@somwehere % rm -rf /var/webconsole /usr/share/webconsole /etc/webconsole /usr/lib/webconsole

Or tarball up and then remove:

root@somewhere % tar cvf /tmp/webconsole.tar /var/webconsole /usr/share/webconsole /etc/webconsole /usr/lib/webconsole
root@somewhere % rm -rf /var/webconsole /usr/share/webconsole /etc/webconsole /usr/lib/webconsole

2. Re-install the WebConsole packages:
If you have your original install DVD/CD handy, then just chuck it in, and install the packages from the Product directory:

root@somewhere % pkgadd -d /cdrom/cdrom0/s2/Solaris_10/Product SUNWzfsgu SUNWmcosx SUNWmcos SUNWasac SUNWmcon SUNWzfsgr SUNWmconr SUNWmctag
The following packages are available:
 1  SUNWasac      Sun Java System Application Server Admin Client
 (i386) 8.2,REV=2006.08.16.08.46
 2  SUNWmcon      Sun Java(TM) Web Console 3.0.2 (Core)
 (i386) 3.0.2,REV=2006.12.08.20.48
 3  SUNWmconr     Sun Java(TM) Web Console 3.0.2 (Root)
 (i386) 3.0.2,REV=2006.12.08.23.39
 4  SUNWmcos      Implementation of Sun Java(TM) Web Console (3.0.2) services
 (i386) 3.0.2,REV=2006.12.08.23.39
 5  SUNWmcosx     Implementation of Sun Java(TM) Web Console (3.0.2) services
 (i386) 3.0.2,REV=2006.12.08.23.39
 6  SUNWmctag     Sun Java(TM) Web Console 3.0.2 (Tags & Components)
 (i386) 3.0.2,REV=2006.12.08.20.48
 7  SUNWzfsgr     ZFS Administration for Sun Java(TM) Web Console (Root)
 (i386) 1.0,REV=2006.10.24.20.44
 8  SUNWzfsgu     ZFS Administration for Sun Java(TM) Web Console (usr)
 (i386) 1.0,REV=2006.10.24.20.44

Even though you are installing version 3.1 of the Java Web Console, it actually appears as 3.0.2 for some stupid reason.

Everything will startup as expected using the new hostname defined.

root@somewhere % wcadmin list

Deployed web applications (application name, context name, status):

console  ROOT            [running]
console  com_sun_web_ui  [running]
console  console         [running]
console  manager         [running]
console  zfs             [running]

Registered jar files (application name, identifier, path):

console  audit_jar     /usr/lib/audit/Audit.jar
console  console_jars  /usr/share/webconsole/lib/*.jar
console  jato_jar      /usr/share/lib/jato/jato.jar
console  javahelp_jar  /usr/jdk/packages/javax.help-2.0/lib/*.jar
console  shared_jars   /usr/share/webconsole/private/container/shared/lib/*.jar

Registered login modules (application name, service name, identifier):

console  ConsoleLogin  userlogin
console  ConsoleLogin  rolelogin

Persistent Jvm options:

-server
-Xmx128m
-XX:+UseParallelGC
-XX:ParallelGCThreads=4

No related posts.

% pkg install gnome

Or

% pkg remove hugin

Or

% pkg verify firefox

Or

% pkg list

I mean.… Why not Sun?

(And what about my cat?)

No related posts.

So, this howto shows you how to run Opera Mini on Linux in glorious 1440x900.… and take advantage of the speed increase.

Requirements

The Linux distribution I run is Debian, but this HOWTO will apply to any distribution really. In fact it should work on Windows as well. All you need initially is a running version of Java. However the complete list is:
Java 6
Micro Emulator
Opera Mini

Installing Java

On Debian I just do the usual installation process:

root@localhost % apt-get install sun-java6-bin sun-java6-jre

That’s it for java.

Installing MicroEmulator

Now for MicroEmulator. Fetch whatever the latest release is . I used 2.0.2 for this setup.

Extract the tarball to somewhere. Doesn’t matter where. I put it under $HOME/lib

mick@localhost [~/lib] # tar zxvf microemulator-2.0.2.tar.gz

I put in a symlink so that I can reference different version if need be. But you can skip this step if you want.

mick@localhost [~/lib] # ln -s microemulator-2.0.2 microemulator

Now, fire up java and point it to the microemulator jar file.

mick@localhost [~/lib] # java -jar $HOME/lib/microemulator/microemulator.jar

This window will now pop-up.

Now Click on ‘Options’, and then ‘Select device’. There’s nothing stopping you from running Opera in this virtual device. But remember you want to run it in a big window, right!

The window below will pop-up.

Click on Add… You want to traverse to your {directory where you installed the microemulator}/devices/, and select the microemu-device-resizable.jar file.

It will now appear in your virtual devices list. Click on ‘Set as default’ to use it every time you start it up.

The MicroEmulator window will change to a ‘not as snazzy’ window like this:

That’s it for the microemulator.

Installing Opera Mini

Installing Opera Mini is easy too. There’s two ways to do it.
Click here , and here and save somewhere.

Or follow these instructions, (I’ve put them here, because, as you know, links change on the internet, and I hate 404s).

Just fetch the latest version from Opera. There’s 4 steps to download:
1. Select ‘Download to PC’.
2. Select ‘Generic version’.
3. Select ‘Advanced MIDP 2′
4. Select you preferred language, and download the JAD and JAR files.

I placed these two files under $HOME/lib/opera. You will have to create a symlink so that the version you have downloaded will be used properly.

mick@localhost [~/lib/opera] # ln -s opera-mini-latest-advanced-int.jar opera-mini-4.1.11355-advanced-int.jar

Your version may/will vary. You’ll need to symlink, (or rename), the file for it to work, though.

Now you want to open the Opera Mini JAD file you just downloaded.

Find it and open it.

Now you will see it in the list of applications that can be launched by Micro Emulator.

Now just double click on it.

Hooley Dooley Batman! It works! Cool!

Either use your mouse and click, or you can use up/down/left/right keys and enter to traverse web pages.

One last thing

I created the following script to make it easier for me to startup, and called it ‘operamini’.

#!/bin/bash

exec >& /tmp/operamini.log

java -jar $HOME/lib/microemulator/microemulator.jar $HOME/lib/opera/opera-mini-latest-advanced-int.jad

Now I have my super fast GPRS internet connection. Also, a side benefit is that all my bookmarks are synced up to my phone.

No related posts.

You see it lacks the thermal control. I’ve been mucking around for a while trying to get the thermal issues under control on the cheap. That is; making use of old power supply fans to keep the air circulating, pulling in cool air, and pushing out warm air. In winter it’s really not a problem. I used to have the door ajar, (and getting dust in at
the same time), and a fan pulling air into the house, which actually warmed up the house a bit. Even in winter, with the door closed and all fans turned off, the cupboard gets to 35C in around 30 minutes. Summer is much worse. With temperatures soaring to 40C, any device will cark it quickly. I’ve had several hard disks die this way.

It’s winter here at the moment, so decided to do something about it. Winter is the easy part. Just suck in cold air from the outside. We’re seeing between 5C to 10C at night so sucking that air in will cool that cupboard down well. I trotted off to Bunnings and bought an inline fan unit that’s designed to pull air from one room and distribute it to
another. It was a bit steep, IMO, at $80, but is just the ticket.

I had to, though, run a large diameter pipe, or tubing from the side of the house which is 6m away. I could have cut into the roof sheeting, but didn’t like compromising the roof, and I wanted to pull in the air from the south side of the house. We fortunately had some left over 100mm PVC drain pipe, so I cut a hole in the side of the house, shoved a 6m length of pipe in, and connected it up with 100mm elbows, to inside the cupboard. I put a filter on the inlet on the side of the house, (hopefully it won’t block up quickly).

The fan is a 150mm fan, so I had to buy a 100mm to 150mm PVC expander. On the end of that I gaffed some springy tubing to the inline fan, and attached some more tubing off the other side. There’s enough there to direct the cold air directly at the air inlets of all the equipment.

I had to dangle the fan from some coat hangers from the ceiling to avoid noise from vibration. I quickly discovered that a 40W fan can generate some loud vibrations. Note the use of coat hangers which are the second best hacking tools, (which I’m sure MacGyver used more than once).

The results were pretty amazing. Below are some RRD graphs of one of my servers, with 4 1Terabyte disks in it. Already you can see a 10C drop:

Next thing to sort out is the summer cooling.

The handy thing with this inline fan is that it has a trap door to stop back drafts. What I might do is, (sort of), the reverse in summer. That is: suck air from the inside of the house, through the cupboard, and into the roof cavity. The cooler air being forced into that space will disperse the hotter air already there, and will consequently make the house cooler. So, I’ll probably need another one of these inline fan units, or else make a valve attachment.

No related posts.

What would you say to a device that can project a 2048x1280 screen on
any surface, infinite focus, no moving parts, that consumes only 1.5W
of power, and is the size of a matchbox?

This device seems to be such a device. If it was a typical 1st April day post, then I would have casually laughed and moved on. But if this device does what it promises, then laser projectors will be the demise of these bulky LCD screens, (LOL).

If only it wasn’t announced a year ago.…. Where is it now?

Read on for more.….

No related posts.

You will have to excuse the quality of these photos — they were all taken with my Nokia camera phone and stitched together afterwards. The quality is shocking, but at least you get the idea.

Laying cables?

A cable laying ship’s primary job is laying fibre cables in the ocean. When you type ‘google.com.au’ most likely, (if you’re not in the US), your traffic is passed along these cables via fibre. Voice calls are also carried over it. A whole tonne of traffic is carried over these little fibres.

Ile de Sein

This ship is big, but not massive. It’s gross tonnage is 13,978, and has the capacity to store 5000 tonnes of cable.

ile de sein

Docked in Sydney

Docked in Sydney

Cable bins

Cables are stored in massive cable bins. The Lle de Sein has two bins, as pictured below. These two bins had deep water cable and shallow water cable.

Deep water cable bin

Shallow water cable bin

The difference between the two types of cable is that the shallow cable needs to be armoured and strengthened to avoid ships scraping the cable, whereas the deep sea cable just needs to protect the fibre cables.

The ship lays the deep water cable at around 10km/h. The cable will finally rest on the ocean floor some 20km behind the ship. Shallow water cable is layed at 500m/h. The reason for the slow speed is that at shallow depths they have to bury the cable to at least a metre below the ocean floor. To do this they use the.…

Trencher

Which is just a massive plough that they drag behind the ship and it lays the cable in the trench as it goes. The trencher depth can be controlled from the bridge.

Trenching tool — side on

The leading edge, which is on the left hand side of this photo has replaceable titanium blades. Apparently after a long haul the blades come up smooth and shiny and as sharp as a razor blade.

Trenching tool — closeup

Trenching tool — from it’s rear.

Roller feeders

Whether it’s deep water or shallow water cable that’s being layed; the laying speed has to be controlled accurately. To do this the cable is brought out of the bins .…

Cable run — from the cable bins

.… and fed through a series of rollers. In the background of the following picture you can see the series of rollers that feed the cable out. The foreground is the roller drum which is used during cable repair.

Cable feeders — laying feeders

Cable feeders — repair feeder

Cable feeders — repair feeder

When cable has to be repaired, the ship will sever the cable on the ocean floor, pick up the cable and move back 6km. The reason is that the cable doesn’t have enough slack to pull it all the way up from the ocean floor, (5000km down). Cable to be repaired will be wrapped several times around the big drum and hauled from the ocean floor. A join will then be made on the cable.

Repeaters

Repeaters have the job of amplifying the signal every so often. There is usually a repeater every 75km.

Repeater A/C unit

Unfortunately I didn’t get a chance to take a photo of one of them. But I took a photo of the A/C unit. The repeaters have to be kept cool and at a constant temperature. The reason is that temperature changes can affect them — and when you’re about to literally sink $100k for each repeater you want to look after it. The seabed is roughly 2C.

Each repeater also has to be powered which by a 5kV DC source. When the ship is laying cable they actually power up the cable and run the repeaters. This is so they can check the quality of the layed cable and ensure there are no issues.

The Bridge

The Bridge is where all the action is.

The Bridge

Cable control room

These days cable laying ships are fully automated. They have high accuracy GPS that can enable them to lay cables pretty exactly. The ship is kept at a constant speed to ensure the cable doesn’t over-stretch or sag. There are two bridges one facing forwards for general operations, (coming and going into port), and one facing backwards for cable laying.

You can download all these images by clicking on the download image below:

No related posts.

The ‘ERROR: (_app_set_slice) Disk is not selected’ error message.

So, you’re trying to setup some DiskSuite metadevices during the jumpstart installation. You are getting this error because you don’t have the slices properly defined in your JET template. You will need to have something similar to the following defined in your JET template file, (under /opt/SUNWjet/Templates):

base_config_profile_root=512
base_config_profile_swap=16384
base_config_profile_s3_mtpt="/usr"
base_config_profile_s3_size="8192"
base_config_profile_s4_mtpt="/opt"
base_config_profile_s4_size="2048"
base_config_profile_s5_mtpt="/var"
base_config_profile_s5_size="2048"
base_config_profile_s6_mtpt=""
base_config_profile_s6_size=""

base_config_profile_disk_c0t1d0s0_mtpt="free"
base_config_profile_disk_c0t1d0s0_size="512"
base_config_profile_disk_c0t1d0s1_mtpt="free"
base_config_profile_disk_c0t1d0s2_size="8192"
base_config_profile_disk_c0t1d0s3_mtpt="free"
base_config_profile_disk_c0t1d0s3_size="8192"
base_config_profile_disk_c0t1d0s4_mtpt="free"
base_config_profile_disk_c0t1d0s4_size="2048"
base_config_profile_disk_c0t1d0s5_mtpt="free"
base_config_profile_disk_c0t1d0s5_size="2048"
base_config_profile_disk_c0t1d0s6_mtpt=""
base_config_profile_disk_c0t1d0s6_size=""
sds_root_mirror="c0t1d0"
sds_database_locations="c0t0d0s7:3 c0t1d0s7:3"
sds_database_partition="c0t0d0s7:64 c0t1d0s7:64"
sds_device_numbers="/:d100:d0:d10 swap:d101:d1:d11 /usr:d103:d3:d13 /opt:d104:d4:d14 /var:d105:d5:d15"

I had this error when I hadn’t specified the *_mtpt and set it to’free’, I had just left it blank.

Jumpstart directory not found

I had the following issue and it took me a while to figure it out.

Searching for JumpStart directory...
not found
Warning: Could not find matching rule in rules.ok
Press the return key for an interactive Solaris install program...

In the end I started to debug the ‘install-solaris’, and ‘profind’ scripts line by line. It turns out that I had defined my hosts with FQDN in my /etc/hosts file. When the server starts to look for the jumpstart directory it’ll use the /usr/sbin/install.d/profind s
cript to determine this. Contained within that is the line:

 set -- `/sbin/bpgetfile -retries 1 install_config`

Because I was using the standard RARP method of booting, bpgetfile was called to determine where to nfs mount the jumpstart directory. bootparams didn’t return any information from the boot server, because it was asking for ‘hostname’, rather than ‘hostname.domain.name’. So, if you get this issue. run the command

/sbin/bpgetfile -retries 1 install_config

from your install shell.

If you get nothing, then this is your issue. You will need to ensure you define your hosts not with a FQDN.

Actually, in my travels on this issue. Most of the resolutions to ‘no matching rule found in rules.ok’ issues had to do with accessing the jumpstart server from the client. So, check all your network settings, (netmasks, routes, etc). If not, then the above resolution will be it. On my clients, this was all OK, so it was baffling for a while.

Also, make sure that you have the matching swap and root partitions set correctly:

base_config_profile_root=512
base_config_profile_disk_c0t1d0s0_mtpt="free"
base_config_profile_disk_c0t1d0s0_size="512"
base_config_profile_swap=8192
base_config_profile_disk_c0t1d0s1_mtpt="free"
base_config_profile_disk_c0t1d0s1_size="8192"

No related posts.