[prev in list] [next in list] [prev in thread] [next in thread] 

List:       evms-devel
Subject:    Re: [Evms-devel] seg fault after install.
From:       greg <greg () fqdn ! com>
Date:       2004-11-13 2:02:15
Message-ID: 41956B27.7010004 () fqdn ! com
[Download RAW message or body]

Thank you Steve for your help and time on this.

It is time to reinstall another vendors linux I'd say. SuSE is known to 
work well with EVMS, and I use it at home. :) Thank you again for 
looking at this, I hope you were able to learn from it as well.

Also thank you Jure for flagging the RHAS issue. To answer your question 
as to why we use RH, 99% of the servers and render farm machines, which 
is hundreds of them, were running it when I took this position. Also 
many of the commercial tools we use are only QA'd on RH. It seemed like 
the sane thing to do, at the time.

take care folks,
greg

>I didn't see any of the new shutdown_thread debug statements in the log.
>That means that either a) you didn't apply and install the patch, which I'm
>sure you did, or b) the code never got run.
>
>This is beginning to sound like the problems with RHAS 3 that Jure Pečar
>mentioned in an earlier post.  The EVMS Engine uses ptheread_create() to
>launch the shutdown_thread().  If the segfault occurs after the call to
>clone() (which I would guess pthread_create() does) as Jure reported, then
>that would explain why there are no debug messages from the
>shutdown_thread() in the log.  It never got run.
>
><take a break>
>
></take a break>
>
>Thanks for the access to your machine.  I tried running gdb and got the
>same results that you did about "Error while reading shared library
>symbols:".  So I tried a different approach.  I edited the Engine code
>(engine/engine.c) and commented out the line that did the pthread_create()
>of the shutdown_thread.  I ran "evms" by itself (without the debugger) and
>it started successfully.  evmsn started successfully, too.  That makes me
>strongly suspect something is wrong in the implementation of
>pthread_create().
>
>I don't have access to the pthread code used in RHAS 3, so I won't be of
>much further help.
>
>I have left the edited engine/engine.c in place in case you want to do
>further testing/evaluation.  The original file is saved as
>engine/engine.c.backup.  The shutdown thread is mainly used for support in
>a clustered environment.  It enables other instances of the Engine that are
>loading to request that the running instance of the Engine shut itself
>down.  If you are not running in a clustered environment then it is not
>needed.
>
>That said, there are other places in EVMS that use pthread_create().  The
>Engine will launch a new thread as part of its copy service which is used
>by object replace, segment move in the DOS and GPT plug-ins, and PE Move
>and PV Move in the LVM plug-in.  I suspect that if you did any of those
>operations you would see the segfault again.  Also, evmsgui launches a few
>threads to do its job.  I suspect evmsgui will continue to get segfaults.
>
>Hope this helps.  A little, at least.
>
>Steve D.
>
>  
>



-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
_______________________________________________
Evms-devel mailing list
Evms-devel@lists.sourceforge.net
To subscribe/unsubscribe, please visit:
https://lists.sourceforge.net/lists/listinfo/evms-devel

[prev in list] [next in list] [prev in thread] [next in thread]