Victor Ivanov

Results 51 comments of Victor Ivanov

It happened again: ``` 2018/09/17 10:02:27 [error] 19964#19964: *1860873 open() "/u/apps/site_production/current/public/uploads/image/1012019/TOLLE_event_12_september_20_uur.png" failed (2: No such file or directory), client: 173.252.95.10, server: *.site.com, request: "GET /uploads/image/1012019/TOLLE_event_12_september_20_uur.png HTTP/1.1", host: "app.site.com" ker process:...

https://github.com/slact/nchan/blob/13e14221ce3e252f4d7c7c47d2dc8d747ec88ae3/src/util/nchan_output.c#L401-L432 Looks like we want to make sure that if the tag is inactive (`t[i] == -1`) that the `tagactive` is not equal to any value of `i`. This is...

Hey @slact, we had two crashes a couple of days ago. Is there a workaround to disable this assertion or the code that generates `Etag` header? We don't store messages,...

Rebuilt with latest `master` and nginx `stable` Got the same error: ``` ker process: /home/deploy/nginx/nchan/src/util/nchan_output.c:424: msgtag_to_strptr: Assertion `id->tagactive != i' failed. 2018/10/08 07:31:54 [alert] 32259#32259: worker process 32261 exited on...

this is pub endpoint log around crash time: `nchan format`: ``` 10.133.18.90 [08/Oct/2018:07:31:50 +0000] webinar~AK1AfZjoRQcap8Fx3s0iisAz8kMMkSgFGg~admin~PiUDdF5z0FtdB7RNjWcbRw - http 1538983905:0 - 10.133.18.90 [08/Oct/2018:07:31:50 +0000] webinar~xaaUvrQ6Inm9MguYoSk64eJAc1ye-bhOfw~admin~Ni51LWCAkNFqvp8zLb1TQg - http 1538983905:0 - 10.133.18.90 [08/Oct/2018:07:31:51 +0000]...

In this specific case the bug was manifested itself because of the specific log format, it turns out one cannot use arbitrary nchan variables in the log format directive. Can...

@delleceste https://nchan.io/#subscriber-presence says: > Note that the subscribe/unsubscribe hooks are disabled for long-poll and interval-poll clients, because they would trigger these hooks each time they receive a message. Those callbacks...

>Is the first stack trace the initial crash? Nope, there were around 30 ctashes, these are last 3 of them. >Anything suspicious in the error logs Not much, I'll post...

> Can you post the backtrace of the first process that crashed? You should be able to tell by the coredump file creation timestamp. I wish I could, but ```...