Good morning, First of all, excuse my bad english, I'm French. So, I want to use 200 whole-root zones with a mysql 5.1.30 community server on each one. I wrote a perl script to create them automatically. This is what the scripts does:
1.detach a "template" zone 2.snapshot the zfs file system of the "template" zone 3.attach the template zone 4.clone the zfs snapshot to a new file system 5.configure the new zone (zonecfg create -a ...) 6.boot the new zone 7.execute some tweaks to configure mysql server in it etc. 8.set cpushares, capped memory I do these operations 200 times, and at the 150th approximatively, the system hangs a lot ! Even a "w" needs 3 to 10 minutes to respond. Errors like: "cannot fork" occure etc... So I modified the script and add an operation: 9.reboot the zone Then, the system responds well and my 200 zones create normaly. My question is: How is it possible ? It seems that the booted zones don't free their memory correctly until they are rebooted... More technical informations: My machine is a bi-xeon (2xe5400) 32Go fo RAM 6 hard disks: -2 for the system (raid hard by lsi card) -4 for the datas ( 2 zfs mirrors) I'm on solaris 10 (I know it's an opensolaris forum, but people here seem to be very skilled !) I set capped-memory to 128M of ram, and reduced the zfs arc size to 1G... Thanks a lot for your replies, Pierre -- This message posted from opensolaris.org _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
