slact

set `/proc/sys/kernel/core_pattern` to `"core.%h.%e.%t.%s.%p"` ``` echo "core.%h.%e.%t.%s.%p" > /proc/sys/kernel/core_pattern ``` set `working_directory` to somewhere nginx can write to, restart it and test this by sending a worker process a SIGABRT....

I also recommend you set the `error_log` directive at the _root level_ of your nginx config -- that's where you set `worker_processes`, `working_directory`, and `events`. set it to level ``notice`:...

> Meanwhile, do the stack traces give you any clues what was going on with nginx/nchan pair? It's probably got something to do with the publisher configuration ``` nchan_message_buffer_length 0;...

not enough to conclusively diagnose the issue. I still need the dump of the first crash after nginx is cleanly stopped and started, ideally with he accompanying error log for...

I'm unable to reproduce this. Does this happen with and without Redis? I suspect this is related to the fix I referenced in #451 . Please check if this happens...

This was a bug in an older version of Nchan that I had been monitoring with that error message. It's unrelated to any of the above settings. It has been...

I see, then this is something different. Can you show me your Nginx config? (edit out whatever you need to keep private)

Thanks for the config. I'll be looking into this. What do you mean by "memory issues"? Also, how often are you getting the "multimsg timeout!" error?

@michaeltcoelho : The issue you're experiencing is different, you have a worker crash which triggers those "multimsg timeout!!" errors. The real problem is ``` nginx: worker process: /nchan-1.1.1/src/util/nchan_output.c:375: msgtag_to_strptr: Assertion...

@indrekj these have been fixed in master with d3a65572046562cd0d92af455ff2ec7eaf466cea