0005151: huge CPU load for spring-dedicated

View Issue Details [ Jump to Notes ]

Project

Tags

No tags attached.

Checked infolog.txt for Errors

Attached Files

spring-dedicated.png (226,525 bytes) 2016-03-07 16:39

spring-dedicated.png (226,525 bytes) 2016-03-07 16:39
20160307_182404_DeltaSiegeDry_101.sdfz (67,953 bytes) 2016-03-07 17:17
fixed_cpu.png (123,401 bytes) 2016-03-09 21:25

fixed_cpu.png (123,401 bytes) 2016-03-09 21:25

Relationships

Relationships

Notes
~0016007 Kloot (developer) 2016-03-07 17:06	and without AI's? server code barely changed between 100 and 101 and it still sleeps 5 millisecs per tick, so the only reason CPU use would be higher is if clients are sending more messages.

~0016008 jamerlan (reporter) 2016-03-07 17:10	no ai. ShockWave have an idea: maybe library was upgraded and now 5 millisec is transfered to 5 nano sec or smth?

~0016009 jamerlan (reporter) 2016-03-07 17:16	attached demo + ShockWave sent me this: nanosleep({1, 0}, NULL

~0016010 Kloot (developer) 2016-03-07 17:23	if you can compile the DS you can test that idea quickly by adding a printf here https://github.com/spring/spring/blob/develop/rts/builds/dedicated/main.cpp#L245 or here https://github.com/spring/spring/blob/develop/rts/Net/GameServer.cpp#L2427 (change the 5 to say 100) and counting how many lines per second are written.

~0016016 abma (administrator) 2016-03-07 21:46 Last edited: 2016-03-07 21:54 View 3 revisions	static build of spring-dedicated 101: ldd spring-dedicated linux-vdso.so.1 => (0x00007ffc200b1000) librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007f28ace47000) libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f28acc3f000) libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f28ac937000) libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f28ac567000) /lib64/ld-linux-x86-64.so.2 (0x00005612390e2000) libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f28ac347000) it links to libpthread, that is an error?! local build spring-dedicated also links to libboost_thread.so edit: seems to be pulled in by the crashhandler/headlessstubs

~0016031 Kloot (developer) 2016-03-09 20:15 Last edited: 2016-03-09 21:45 View 2 revisions	note, as of f188aad8b91aca28a39d6378f675b6b008086c95 the server sleeps can be controlled: gameserver now has a ServerSleepTime config parameter (milliseconds, default 5) dedicated now has a --sleeptime commandline parameter (seconds, default 1) test this DS build and report back please.

~0016033 jamerlan (reporter) 2016-03-09 21:36	Looks like now issue is solved. Without additional arguments passed :-) (see fixed_cpu.png) Strange that bug was reproducible only on BlackHoleHost (Debian GNU/Linux 8 (jessie)) but not on ACE hosts (unknown OS) I was unable to reproduce on my own host (CentOS Linux release 7.2.1511) too.

~0016034 Kloot (developer) 2016-03-09 21:44	Can you find out which versions of g++ and libboost are installed on BHH?

~0016035 jamerlan (reporter) 2016-03-09 22:44	blackhole: gcc http://pastebin.com/MNBkH2P1 libboost: http://pastebin.com/CwhrpDnw

~0016036 jamerlan (reporter) 2016-03-09 22:46	testing@BlackHoleHost:~$ dpkg-query -l 'g++*' ?······························································································ Desired=Unknown/Install/Remove/Purge/Hold ?······························································································ \| Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend ?······························································································ \|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad) ?······························································································ \|\|/ Name Version Architecture Description ?······························································································ +++-====================-===============-===============-============================================= ?······························································································ ii g++ 4:4.9.2-2 amd64 GNU C++ compiler ?······························································································ ii g++-4.9 4.9.2-10 amd64 GNU C++ compiler ?······························································································ un g++-4.9-multilib <none> <none> (no description available) ?······························································································ un g++-multilib <none> <none> (no description available)

~0016043 Kloot (developer) 2016-03-11 02:22	there was a bug in 1.55 boost::chrono that would cause this_thread::sleep_for to return immediately, but spring uses std::chrono on linux so no idea besides "system/kernel issue". thanks for testing anyway.

Notes

Date Modified	Username	Field	Change
Issue History
2016-03-07 16:39	jamerlan	New Issue
2016-03-07 16:39	jamerlan	File Added: spring-dedicated.png
2016-03-07 17:06	Kloot	Note Added: 0016007
2016-03-07 17:10	jamerlan	Note Added: 0016008
2016-03-07 17:16	jamerlan	Note Added: 0016009
2016-03-07 17:17	jamerlan	File Added: 20160307_182404_DeltaSiegeDry_101.sdfz
2016-03-07 17:23	Kloot	Note Added: 0016010
2016-03-07 21:42	abma	Target Version	=> 102.0
2016-03-07 21:46	abma	Note Added: 0016016
2016-03-07 21:53	abma	Note Edited: 0016016	View Revisions
2016-03-07 21:54	abma	Note Edited: 0016016	View Revisions
2016-03-09 20:15	Kloot	Note Added: 0016031
2016-03-09 20:15	Kloot	Status	new => feedback
2016-03-09 21:25	jamerlan	File Added: fixed_cpu.png
2016-03-09 21:36	jamerlan	Note Added: 0016033
2016-03-09 21:36	jamerlan	Status	feedback => new
2016-03-09 21:44	Kloot	Note Added: 0016034
2016-03-09 21:45	Kloot	Note Edited: 0016031	View Revisions
2016-03-09 22:44	jamerlan	Note Added: 0016035
2016-03-09 22:46	jamerlan	Note Added: 0016036
2016-03-11 02:22	Kloot	Note Added: 0016043
2016-03-11 02:22	Kloot	Status	new => resolved
2016-03-11 02:22	Kloot	Resolution	open => no change required
2016-03-11 02:22	Kloot	Assigned To	=> Kloot

Issue History