postponed for another 8h at least.

Hi @utzer [Friendica] , are there still troubles with the *.friendi.ca server?

I restarted the friendica.me instance, based on the develop branch and now I'm unable to build it again
- files.friendi.ca is down for the archives
- git.friendi.ca is down for the json-ld library

Unfortunately, I'm now stuck 😕

@Philipp Holzer the server is running again, it crashed, nothing in the log, wr did a run of fsck. Should have done that before.

Now we need to wait again to see what happens. Dmesg is open and we're capturing the output.

Good luck finding the root - cause 🙁 🙁 I know how annoying such crash situations are

@Philipp Holzer @utzer [Friendica] There should be some messages in php-fpm.log or syslog at least. Or do you have #systemd ? Then try reading its journal. A crash, mostly SEGFAULT means that there is an endless/recursive loop. I had this here. Try to reproduce it with strace php crashing-script.php (minimum) and also add a lot debug messages, e.g. which method is being invoked with which parameters (log them all). This is what I do and I was always very successful with it:

// Minimum (sometimes the only way in segfault situations):
printf('[%s:%d]: param1[%s]=%s - CALLED!' . PHP_EOL, __METHOD__, __LINE__, gettype($param1), $param1);

// Or with an actual logger
SomeLogger::trace(sprintf('param1[%s]=%s - CALLED!', gettype($param1), $param1);

Of course logging the type of the parameter makes only sense when you have a multiple-type e.g. string $param1 = null, then $param1 could be type of string or null.

@Roland Häder @Philipp Holzer maybe it was the file system. All logging stopped before the crash, so that looks like read only because of an error.

@utzer [Friendica] @Philipp Holzer Okay, maybe the on-error condition has made it read-only (default for most distris). errors=remount-ro is then the option, it is very highly recommended.