**** BEGIN LOGGING AT Mon Oct 02 03:00:01 2017 Oct 02 10:15:01 ohnoes, load 35 again Oct 02 10:16:10 pabs3: too many trac.fcgi again, load shoots the roof, no free RAM Oct 02 10:16:49 well, the latter is not *really* the problem, it seems it's mostly cache/buffers Oct 02 10:18:29 pabs3: http://susepaste.org/7667687 Oct 02 10:26:54 and since neither http://docs.openmoko.org/trac/ nor e.g. http://wiki.openmoko.org/wiki/Distributions#Images drops requests with a "temporarily overheated", the situation explodes in an avalanche as soon as there's one more request per time window than the server can serve Oct 02 10:28:07 wow, after 2 minutes, http://wiki.openmoko.org/wiki/Distributions#Images browser claims "55% progress on loading" Oct 02 10:29:34 http://susepaste.org/76704260 Oct 02 10:32:10 also *cough* @ /usr/bin/git --git-dir=/space/git/repositories/git/kernel.git rev-list --header --full-history --max-count=101 --skip=100 refs/tags/v2.6.23-rc1 -- fs/ext3 Oct 02 10:36:44 or stuff like usr/bin/git --git-dir=/space/git/repositories/git/kernel.git rev-list --header --author=Alistair John Strachan --regexp-ignore-case --fixed-strings --max-count=101 --skip=0 557e993224b12f6f9d Oct 02 10:39:15 I don't think *anybody* is actively using (as in: "writing") git or particularly trac anymore. Either we shut that stuff down completely or just convert it to read-only static content Oct 02 10:40:11 I'd not be surprised if trac gets massively abused by spambots Oct 02 10:40:21 creating link farms Oct 02 10:49:28 maybe a robots.txt on trac may already help Oct 02 10:51:26 seem obvious to me that something tries to do a wget mirror of the docs.openmoko.org and git website, with infinite depth so e.g. git gets searched for bizarre queries like "--author=Alistair John Strachan" Oct 02 10:52:52 pabs3: your MaxClients helped in that it now seems steady @ load:30 instead of former load:140 Oct 02 10:53:45 still we need a load <8 Oct 02 10:53:55 aiui Oct 02 11:22:51 being a *total* noob on webserver config, I'd prolly consider looking into MaxSpareServers, MaxRequestsPerChild, ThreadsPerChild, ServerLimit, and MaxSpareThreads in addition to MaxChilds Oct 02 11:23:50 and check robots.txt Oct 02 11:59:55 pabs3: please check http://httpd.apache.org/mod_fcgid/mod/mod_fcgid.html Oct 02 12:22:17 particularly http://httpd.apache.org/mod_fcgid/mod/mod_fcgid.html#fcgidmaxprocesses Oct 02 12:22:41 assuming the server uses mod_fcgid Oct 02 20:54:54 pabs3: did you change any settings? Oct 02 20:55:35 right now only 4 *trac.fcgi* processes and basically load:0 Oct 02 20:56:25 5 Oct 02 20:57:52 unrelated SecInfo: https://security.googleblog.com/2017/10/behind-masq-yet-more-dns-and-dhcp.html Oct 03 00:10:42 DocScrutinizer05: no, Harald may have Oct 03 00:11:03 anyway looked good Oct 03 00:11:12 still looks good Oct 03 00:11:36 I wonder if the spiders stopped or some config got changed Oct 03 00:12:21 don't you think we could completely shut down admin-trac.om.org? Oct 03 00:13:21 no idea, I can look later (half awake at this point) Oct 03 00:14:51 np, right now everything is absolutely fine :-) Maybe consider adjusting http://httpd.apache.org/mod_fcgid/mod/mod_fcgid.html#fcgidmaxprocesses later on? I'm not saying it's the right parameter to tweak, just looks somewhat related to me (apache/fcgi/trac noob here) Oct 03 00:16:10 hmm dynamic.kbtelecom.net and wa.optusnet.com.au Oct 03 00:17:39 so yep, most likely Harald? Oct 03 00:18:29 or somebody using a leete VPN ;-) Oct 03 00:21:29 yeah **** ENDING LOGGING AT Tue Oct 03 03:00:00 2017