Valentin Chopov
valentin@valcho.net
Wed Apr 23 14:26:47 UTC 2003
Vernon, no the "cron-dccd" on the backup server is cron'ed for 6:50am Here is the part of "cdcc info" from 04/23/03 03:08:02 ED: 78% of 32 requests ok 2545.87-800 ms RTT 2740 ms queue wait Even now, when the main server is used, the backup server gives: 78% of 32 requests ok 1479.19-800 ms RTT 275 ms queue wait BTW, when this server serves dccm requests everything is fine, like: 100% of 32 requests ok 14.02-800 ms RTT 62 ms queue wait It seems that when the backup server is "sleeping" "cdcc info" gives high results.... Thanks, Val On Wed, 23 Apr 2003, Vernon Schryver wrote: > > From: Valentin Chopov <valentin@valcho.net> > > > ... > > I'm running a backup dccd server and every night at ~3am when the > > "cron-dccd" is running on my main dccd server it's busy but instead of > > switching to my backup server dccm uses one of the public dccd servers. > > The dccd backup server is on the same LAN and the map.txt is something > > like: > > > > dcc.dcc-servers.net,- RTT+1000 ms anon > > maindccd,- RTT-1000 ms clientid pass > > backupdccd,- RTT-800 ms clientid pass > > > > The backup server is not loaded at all and doesn't run "cron-dccd", etc. > > at this time. > > ... > > What does `cdcc info` say at ~2:59 am? Is it possible that the backup > server runs `cron-dccd` earlier in the night so that its measured RTT > and queue delay is larger than 2000 ms? That would make the public > servers look like better choices. > > Some flavors of UNIX don't cope well with grinding 200 MByte files in > dbclean, so that concurrent dccd or any other activity goes very > slowly. BSD/OS through 4.2 is particularly bad about that. > > > Vernon Schryver vjs@rhyolite.com > _______________________________________________ > DCC mailing list DCC@rhyolite.com > http://www.rhyolite.com/mailman/listinfo/dcc > >
More information about the DCC
mailing list