https://bz.apache.org/bugzilla/show_bug.cgi?id=65402
--- Comment #43 from Stefan Eissing <ste...@eissing.org> --- If you don't know what is going on, write a helper. Today I wrote <https://github.com/icing/h2wtf> to look at your logs. My analysis now is that your webapp runs into the h2 DoS protection. mod_http2 does not really like it when the client starts and then kills requests. But it likes when the client reads requests completely and without stalling. There is a "mood" counter and when mood drops, the number of active requests that a client may perform is reduced. When mod_http2 is happy, this limit is increased. The stream with the late starts happen at a time, when the limit has been exhausted. E.g. in the latest log, the limit was reduced to 16 and 16 request were ongoing when a new one came in. This means there will be no worker assigned to that request until another of the 16 returns. Since there are requests hanging, waiting on the next SSE chunk to then abort, this may take some time. In the last log the 16 seconds. So, I understand what is happening. That is something. I do not have a good idea how to improve this. Have to think a bit on that. -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: bugs-unsubscr...@httpd.apache.org For additional commands, e-mail: bugs-h...@httpd.apache.org