XH-H01: The server was stuck: no frame for seconds

On the console:

[XH-H01] Server stuck for 10 s in hangtest.amxx (commands); report xhlds/crashes/<id>.dmp; restart at 60 s

Printed: when the server finished no frame for xhlds_hang_report seconds (Linux), then when it runs again or is ended (Server running again after 23 s stuck ..., Server still stuck after 60 s ...; ending the process to restart it ...).

The server finished no frame for xhlds_hang_report seconds (10 by default; 6 times that while a map loads or the server starts): it was stuck, and every player's game froze. The usual cause is a plugin that does not return: an endless loop, or a very long one (a big file, a database query that waits, a loop over every entity or player many times). The server wrote a crash report from where it was stuck; the line names the AMX Mod X plugin that was running (or else the module, e.g. cs.so+0x1a2b) and what the frame was doing (commands, read_packets, physics, map_load, ...). On hostsrc's crash service a hang counts as a crash (signal HANG). "location unknown: the main thread did not take the signal" means the place could not be found (the report has the other threads); "no report (...)" says why there is none.

If it comes back, a second line says "Server running again after N s stuck"; the report is kept. If it is still stuck after xhlds_hang_restart seconds (60 by default, 3 times that in a map load), the server ends itself ("ending the process to restart it", exit status 124) and hlds_run or the panel starts it again, as after a crash (see XH-H03 for when it does not). Quitting the server is never counted.

Is it a problem? Yes: everyone on the server froze for that long.

What to do:

• Look at the plugin the line names: update it, or take it out of plugins.ini and watch whether the hangs stop. Without a plugin, the module in the line (pc_module= in the report) says whose code it was.
• The report (xhlds/crashes/<id>.txt and .dmp) is uploaded like a crash report; send it to us if not.
• xhlds_hang_report 0 turns the watchdog off; xhlds_hang_restart 0 reports a hang but never ends the server.

On the server

xhlds_explain XH-H01 prints this explanation on the server console, also over rcon.