New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
4.1.3: Varnish crashes with "Assert error in CNT_Request(), cache/cache_req_fsm.c line 820" #2106
Comments
This is a known and fixed issue, see #2000. |
My bad, reopening. |
This looks like an incorrect restart. Can you share the VCL (at least those bits where you're restarting)? Are you using any inline-C or vmods? |
We are have vmod softpurge and variable . The variable vmod is not official is this one : We are using softpurge and standard purge because we have some cache objects that we want to refresh quickly and other don't. We have object variants by user type. Sometimes we override the vary header to make global cache objects (Static files). We are using Drupal 7.50 with Authcache http://pastebin.com/m36S01yQ --> drupal.vcl |
We removed the variable vmod and now the panic is reduced. We have a interval of 7 days for each panic |
So you are seeing the same panic, just not that often? |
@Lutacon because the issue you are seeing is so strange, I think we should exclude other sources of error:
|
First of all. Sorry for the late response we are having a busy month >_< @fgsch Yes. We are seeing the same panic but with an interval of 3-4 days. @nigoroll The answers to your questions :
Updated VCL : http://pastebin.com/AcxtSTrQ Updated panic:
|
Do I understand correctly that
This does not sound good. Please ensure that you (re-)compile your vmods with exactly the same varnish installation you are running and report back |
Yes. It's correct. We updated from 4.1.2 to 4.1.3 and we didn't re-compile vmods after update. Ok. We are going to recompile vmods with version 4.1.3 and i'll report back |
We are going to reinstall varnish and compile vmods. Anyway we had this panic before upgrading to 4.1.3 |
Hey guys, we're seeing the same issue a few times a day here at Twitch. I haven't really started to dig into it but we're running 4.1.3. I can try to gather more info if it would be helpful. |
In 4.1.3 RESTART is mapped to 0 so this is not a restart but most likely req not properly (re)initialized. Do you know or could you capture if these requests are being put in the waiting list? |
Hello, An updated about our situation. We had in our VCL the vmod STD and we were logging some requests. We didn't include in the bug report because we thought it wasn't necessary... We have removed the STD vmod and we didn't have a crash since 23/10. We are going to try the same configuration on our development server with STD to try to reproduce it. |
Hello, Same bug here with exactly same Varnish version on Debian Wheezy.
We dont use "restart" but "retry" in our VCL. Tell me if you need more information |
@llavaud We disabled the std vmod and we didn't have a crash since 23/10. You could try that. |
ok great to know there is a workaround but std vmod is very helpful... :( |
@Lutacon I checked your VCL and I don't see where you are/were using std. |
@fgsch When we post our vcl we did some cleanup to remove some internal comments and our IPs. We also removed the std logs. This is where we were using std :
|
@Lutacon thanks. So having those 2 lines is enough to cause the crash? |
@Lutacon also any luck trying to reproduce it? |
@fgsch We were not be able to reproduce it on our development server with the same setup but we though it's because we don't have the same traffic. And since we remove those 2 lines we didn't have a crash. varnish param.show output:
|
@Lutacon OK. Do you happen to have the varnishstat output at the time of the crash? |
@fgsch I don't have the varnishstat output for that time. About the thread_pool setting we are thinking about removing it from the daemon but we saw those parameters from varnish book. At this time we have this in the daemon options:
We can't try that at this moment because of the black friday and ciber monday. If we have a crash on those dates we are fucked. I'll let you know when we try those options. |
@Lutacon you can remove it entirely. Where in the Varnish Book this is? |
@mbgrydeland I will fix the changelog |
Backport review: This is fixed in the latest version of 4.1, and @mbgrydeland also fixed the corresponding problem in master. |
Seems that the issue is still there on varnish-4.1.4. Root cause though has moved from line 820 to line 824: After upgrade: Feb 13 05:12:14 hostname varnishd[6804]: Child (14618) not responding to CLI, killed it. OS: Packages: Let me know if there is any more debug information I need to add. Thanks, |
The fix is in 4.1.5, not 4.1.4 |
Thanks @scoof. Have upgraded to 4.1.5 now. Fingers crossed. :) |
@jansonz Did you have any more crashes after upgrading to 4.1.5? |
Hi there,
We have this problem each day. Varnish crashes and we lost everything we have in the cache. We thought we could fix this problem upgrading to 4.1.3 from 4.1.2 but the problem persists.
Your Environment
The config in varnish.etc :
The text was updated successfully, but these errors were encountered: