Stefan Eissing

There is a known issue since the dawn of time in restart/reload behaviour and watchdog tasks. The problem is that the order in which watchdog tasks and server shut down...

The first stack trace shows ``` #2 0xf3cce598 in CRYPTO_THREAD_write_lock (lock=0x0) at crypto/threads_pthread.c:78 ``` and that means the lock pointer is NULL which is, afaict not supposed to happen. Openssl...

Hmm, looking at the differences between 2.4.53 and 2.4.54, there was a change in JSON reference handling. Can you try the following patch? ``` Index: modules/md/md_json.c =================================================================== --- modules/md/md_json.c (Revision...

Well, that means the latest changes are not causing the issues on your system. Have you switched/updated other components? Maybe from OpenSSL 1.1 to 3.0? Trying to circle things down...

Ok, scanning the code and the report again. Can you produce a log at `LogLevel md:trace4` of what happens right before the crash? You can mail it to me at...

@whereisaaron I am investigating the issue with @arminabf and so far, he is the only one seeing it. I proposed a fix in the current master that restores the old...

@arminabf confirmed the issue to be fixed in the last patch which I now release as v2.4.19.

The APR sales rep, who sold me this feature, never mentioned this!

Thanks for the analysis and logs. Indeed, the question seems to be why the stream output beam has been aborted before any response was produced. Do you see any lines...

Thanks for digging these up! > however, would this mean that we need to increase H2MaxWorkers or H2MaxSessionStreams? im sure we tried out increasing H2MaxWorkers and it didn't help... No,...