Headless Linux squeak running max cpu

I start the image with the following command:

setsid bin/squeak -vm-display-null -vm-sound-null echat-server.image introducer.sq &

which runs it as a daemon, turns off display and sound, and runs a small script that starts a listening socket.  The output from top, after running a few minutes, is

 PID       USER      PR  NI  VIRT     RES  SHR   S  %CPU %MEM    TIME+ COMMAND
12557 vawhigso  20   0 1036m  42m 1276  S    23.4      0.3      0:28.32 squeak

There are 8 cpus, so we are getting time on more than one cpu.  Memory has grown from 0.2 to 0.3 but doesn't seem to be growing ridiculously.  At some point in time (several hours, I believe), squeak is no longer running.  I do not know if this is because it runs out of memory, some other internal exit, or the webhost kills it.   It looks like it is infinitely looping, but not grabbing memory.   I expect it to idle at extremely low cpu.

What kinds of debugging tools can I use to see what it is doing?

Thanks,
Rob

I start the image with the following command:

setsid bin/squeak -vm-display-null -vm-sound-null echat-server.image introducer.sq &

which runs it as a daemon, turns off display and sound, and runs a small script that starts a listening socket.  The output from top, after running a few minutes, is

 PID       USER      PR  NI  VIRT     RES  SHR   S  %CPU %MEM    TIME+ COMMAND
12557 vawhigso  20   0 1036m  42m 1276  S    23.4      0.3      0:28.32 squeak

There are 8 cpus, so we are getting time on more than one cpu.  Memory has grown from 0.2 to 0.3 but doesn't seem to be growing ridiculously.  At some point in time (several hours, I believe), squeak is no longer running.  I do not know if this is because it runs out of memory, some other internal exit, or the webhost kills it.   It looks like it is infinitely looping, but not grabbing memory.   I expect it to idle at extremely low cpu.

What kinds of debugging tools can I use to see what it is doing?

Thanks,
Rob

It is with my new Cog binary built with your help for a Linux server. 

 

Are you saying I can send SIGUSR1 to squeak to make it do that?   How do I send SIGUSR1 again to a process?

 

Thanks,
Rob

From: [hidden email]

Sent: Sunday, July 18, 2010 3:24 PM

To: [hidden email]

Subject: Re: [squeak-dev] Headless Linux squeak running max cpu

On Sun, Jul 18, 2010 at 9:56 AM, Rob Withers <[hidden email]> wrote:

I start the image with the following command:

setsid bin/squeak -vm-display-null -vm-sound-null echat-server.image introducer.sq &

which runs it as a daemon, turns off display and sound, and runs a small script that starts a listening socket.  The output from top, after running a few minutes, is

 PID       USER      PR  NI  VIRT     RES  SHR   S  %CPU %MEM    TIME+ COMMAND
12557 vawhigso  20   0 1036m  42m 1276  S    23.4      0.3      0:28.32 squeak

There are 8 cpus, so we are getting time on more than one cpu.  Memory has grown from 0.2 to 0.3 but doesn't seem to be growing ridiculously.  At some point in time (several hours, I believe), squeak is no longer running.  I do not know if this is because it runs out of memory, some other internal exit, or the webhost kills it.   It looks like it is infinitely looping, but not grabbing memory.   I expect it to idle at extremely low cpu.

What kinds of debugging tools can I use to see what it is doing?

In Cog platforms/unix/vm/sqUnixMain.c you'll find code to cause SIGUSR1 to print all stacks.  This can easily be ported to the standard VM. Is the above with Cog or the standard VM?

cheers

Eliot

Thanks,
Rob

---

It is with my new Cog binary built with your help for a Linux server. 

 

Are you saying I can send SIGUSR1 to squeak to make it do that?   How do I send SIGUSR1 again to a process?

 

Thanks,
Rob

From: [hidden email]

Sent: Sunday, July 18, 2010 3:24 PM

To: [hidden email]

Subject: Re: [squeak-dev] Headless Linux squeak running max cpu

On Sun, Jul 18, 2010 at 9:56 AM, Rob Withers <[hidden email]> wrote:

I start the image with the following command:

setsid bin/squeak -vm-display-null -vm-sound-null echat-server.image introducer.sq &

which runs it as a daemon, turns off display and sound, and runs a small script that starts a listening socket.  The output from top, after running a few minutes, is

 PID       USER      PR  NI  VIRT     RES  SHR   S  %CPU %MEM    TIME+ COMMAND
12557 vawhigso  20   0 1036m  42m 1276  S    23.4      0.3      0:28.32 squeak

There are 8 cpus, so we are getting time on more than one cpu.  Memory has grown from 0.2 to 0.3 but doesn't seem to be growing ridiculously.  At some point in time (several hours, I believe), squeak is no longer running.  I do not know if this is because it runs out of memory, some other internal exit, or the webhost kills it.   It looks like it is infinitely looping, but not grabbing memory.   I expect it to idle at extremely low cpu.

What kinds of debugging tools can I use to see what it is doing?

In Cog platforms/unix/vm/sqUnixMain.c you'll find code to cause SIGUSR1 to print all stacks.  This can easily be ported to the standard VM. Is the above with Cog or the standard VM?

cheers

Eliot

Thanks,
Rob

---

--------------------------------------------------
From: "Levente Uzonyi" <[hidden email]>

Sent: Sunday, July 18, 2010 3:42 PM

To: "The general-purpose Squeak developers list" <[hidden email]>

Subject: Re: [squeak-dev] Headless Linux squeak running max cpu

Install RFB (it's available on squeaksource). It will give full access to the image. If you never used it, here are a few tips:
After installing the package, in the world menu select open..., then RFB/VNC Server. In the connections submenu set an interactive password. It doesn't have to be strong, since you'll tunnel your connection through ssh. In the connections submenu disable "allow remote connections". In the performance submenu set "allow ZRLE encoding" and disable "conserve memory". Then start your server. Note that the tcp port is display number + 5900. So if you want your server to listen on 5901, then select display number 1.

Squeak's RFB implementation is a bit picky, it won't accept connections with certain client settings. For example it doesn't support other display bitdepth than the image's depth. So if you're using RealVNC as a client, then set the preferred encoding to ZRLE and select Full color (your image probably uses 32-bits as depth). And make sure you turn off the "Render cursor locally" option.

What does RFB stand for?  I intend to check this out.

Thanks,
Rob