Squid 2.5.9 without customlog patch was working perfectly. Now after upgrade to 2.5.9-r2 it shows in logs: 2005/04/18 23:56:51| assertion failed: stmem.c:93: "current_offset == target_offset" And then starts over again. So every few connections users get "Connection refused", when squid is starting again.
wow! I've made customlog flag because of your complaint about memory leaks. I'm beginning to think that you have serious problems with your hardware (memory failure perhaps?). I bet you are the only one experiencing such assertions :)
I have scanned my logs and see nothing like that. May be this is some kind of problem with that system/hardware? Do you have especially high load, unusual hardware or anything of that nature?
I have load on that machine near 2.0. It is 2 x Xeon 2.66. I am sure that there are no memory problems because more software would fail and that machine was working months without such failures. Squid was also working fine, but when upgrading to 2.5.9-r2 it starts to fail right away (in no more than 30 seconds). Maybe it is due to high load, hard to say. Tried removing its spool, but it didn't help. One thing is sure - pure 2.5.9 is working flawlessy. Pure I mean that from portage with version 2.5.9 and customlog disabled. I guess that you don't get theese trouble because of lower load, less customers or something like that (plenty machines of my customers are hamerring with FrontPage exploits and it goes via proxy because it is transparent).
And also: bash-2.05b# grep stmem.c * | cut -d : -f 1 | sort | uniq 15_all_squid-2.5.STABLE9-aufs.patch 21_all_squid-2.5.STABLE9-2GB.patch So stmem.c is being modified in emerge process.
I may have low load on my servers, but how do you explain that you are the only one having problems with customlog and 2.5.9-r2? Are you saying are the only one gentooer with high load on their squid? Please run memtest on your machine. comment #4 : these patches are official and it will be included in the next release of squid
same thing happends to me also, the prolem apears when squid is configured with diskd, the same problem is already discussed here: http://www.mail-archive.com/squid-users@squid-cache.org/msg27578.html read the thread, there is a patch there, which i didn't apply. i switched to ufs until squid will be fixed in portage.
it seems there is a problem with the 2GB patch.
So there is a bug as you may see. And once again - my hardware is not faulty (I would know that because I have machine with IPMI support, and I can see even if there are memory errors in System Event Log), and wasn't faulty with customlog patch. Some people need to understand that squid is quite huge project, it is configurable with hundreds of options, and it is impossible for anybody to check every possible option. So if squid works fine for you and hundred of other gentooers doesn't mean it will work for 101th gentooer.
You also need to understand that we are only human, and need to be able to reproduce bugs or look to upstream for solutions. We are already well aware of what a big project squid is, and I think do our best to solve any issues which arise. It is of course very difficult when you are unable to reproduce the bugs being described and can see no clear reason for them. One possibility was also that your hardware could have been faulty, and it is not as if this has not happened in the past. You shouldn't feel insulted that it was asked, as Alin was just trying to work through possible causes with hardware issues sometimes producing some very strange bugs. I myself had a segfaulting tar for tars over 3GB, despite the system passing memtest86 for 8 hours...
I understand that of course and I am glad that you are doing such a great work. Before installing gentoo I was part-time developer of small polish distribution and I know well what "pain in the ass" is debuging a problem. But I can assure you that when I head to bugs.gentoo.org to fill a bugreport I am quite sure that I checked everything else. Anyway can't wait for 2.5.9-r3 with fixed 2GB patch :D.
Just wanted to let you know, that I see the same problem on two fujitsu-siemens RX200 with 4GB, running on high load. I already searched for a few days, can't do further tests, downgrading.
fixed in cvs by r3. I had to rebuild the entire patch tarball, just for being sure. please tell me if it solved the problem. I'm on thin ice here as I've broken gentoo policy about marking packages as stable :(
assert message is gone, everything seems to work fine so far. Thanks for the quick fix.
In my case it is doing segfaults from time to time. Going back to 2.5.9, sorry. Apr 20 00:30:50 [squid] Squid Parent: child process 22541 exited due to signal 11 Apr 20 00:30:53 [squid] Squid Parent: child process 23043 started
declaring this as fixed since the original problem is solved. @Michal: please monitor logs to see whether 2.5.9 has the same behaviour or not
2.5.9-r3 now works clean for me, not one single problem within the last 12 hours.
