[Seaside] VM stability on UNIX

Dmitry Dorofeev dima-sender-3c337a at yasp.com
Tue Oct 25 09:47:06 UTC 2005



Adrian Lienhard wrote:
> Hi Dmitry,
> 
> We've had problems initially with Postgres and Sockets on Debian,  
> though they resulted in different errors. The solution was setting  the 
> -notimer argument for the vm.

Nice hint, as I gonna use PostgreSQL soon.

> 
> Another source of problem was leaving open the Transcript window,  which 
> quite fast ended up locking the image (IIRC), although  Transcript is 
> claimed to be thread-save... Since you run headless,  this should not be 
> a problem, though.
Yep, I have the Transcript opened. But it is so in all 4 images, but only one
is freezing.

> 
> Then, there is the weak WeakArray implementation. Oh, Lukas just  wrote 
> about this... ;-)

Hmm, It is not any load to the apps, as they all in a test mode. I.e.
only me and couple of developers accessing it. Plus other 3 still running fine
even with higher load (some client clicks)

> 
> To get more information you could maybe hook into the vm with gdb and  
> get a stack trace?

squeak was compiled without debug, so not much there.
I sent kill -STOP to a bugger yesterday, and today i attached gdb.

backtrace has 15749 items! So it is clearly dead loop somewhere:

#15748 0xcfbef2f4 in ?? ()
#15749 0x00000009 in ?? ()

Cannot access memory at address 0xcfbfe000


> 
> I would also try running without Squeak's RFB instead to get rid of  one 
> possible source of problems. I guess, that "select: Invalid  argument" 
> could come from there...

Also other 3 apps feels quite healthy with RFB

-Dmitry.

> 
> Adrian
> 
> 
> On Oct 24, 2005, at 3:08 PM, Dmitry Dorofeev wrote:
> 
>> Hi all,
>>
>> I have 4 squeak VM running on the same UNIX (OpenBSD) server.
>> All running separate and different Seaside apps in a test mode (yet).
>>
>> First trick is to start squeak in a 'safe' way. I.e. you must use
>> squeak -nodisplay -nosound -memory 64m your.image
>>
>> If you dont limit memory usage, it may eat 100% of CPU and stop  
>> responding.
>> Which is known UNIX VM bug as I get from the lists. Its related to  
>> memory usage
>> over 1Gb.
>>
>> Another problem which I have no solution for is one my squeak  process 
>> eats
>> about 90% of CPU time to time. If I wait enough (several hours) it  
>> may go normal.
>> It prints to STDOUT millions of messages:
>>
>> select: Invalid argument
>>
>> It is the only app which uses MySQL. But I close MySQL connection  
>> after SQL request done. So unsure if it is related to the problem.  
>> Also it is app which sends emails with:
>>
>> SMTPClient
>>     deliverMailFrom: m fromEmail
>>     to: (Array with: e)
>>     text: data
>>     usingServer: 'localhost'.
>>
>> All squeaks run RFB (VNC) on a separte ports.
>>
>> Squeak version i use:
>> # squeak -version
>> 3.7-7 #2 Sat Sep  3 17:14:42 BST 2005 gcc 3.3.5
>> Squeak3.7 of '4 September 2004' [latest update: #5989]
>> OpenBSD emily.flightstore.co.uk 3.7 EMILY#5 i386
>>
>> Does anyone have similar problem with message 'select: Invalid  argument'
>> or have any clue ?
>>
>> Thanks,
>> -Dmitry Dorofeev.
>> _______________________________________________
>> Seaside mailing list
>> Seaside at lists.squeakfoundation.org
>> http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside
>>
> 
> _______________________________________________
> Seaside mailing list
> Seaside at lists.squeakfoundation.org
> http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside


More information about the Seaside mailing list