2014-05-02 | 0 comments
[root@foonsd3 ~]# mmlsconfig minReleaseLevel minReleaseLevel 3.5.0.3 [root@foonsd3 ~]# mmchconfig release=LATEST Verifying that all nodes in the cluster are up-to-date ... mmchconfig: Command successfully completed mmchconfig: Propagating the cluster configuration data to all affected nodes. This is an asynchronous process. [root@foonsd3 ~]# mmlsconfig minReleaseLevel minReleaseLevel 3.5.0.11
2014-04-13 | 0 comments
Here is an example of demoting two GPFS servers named foonsd1 and foonsd2 into being regular GPFS clients. This means removing them from being part of cluster “quorum”, declaring that it is to be a client node (ie, can not be a token manager), designating new configuration servers, and changing the licensing terms for those nodes. Three new GPFS servers have already been added to the cluster so that at no point is quorum lost.
In order to change the quorum status of a node, GPFS must not be running.
# mmshutdown -N foonsd1,foonsd2 Sun Mar 23 14:20:38 MST 2014: mmshutdown: Starting force unmount of GPFS file systems Sun Mar 23 14:20:43 MST 2014: mmshutdown: Shutting down GPFS daemons foonsd2.example.org: Shutting down! foonsd1.example.org: Shutting down! foonsd2.example.org: 'shutdown' command about to kill process 3256 foonsd2.example.org: Unloading modules from /lib/modules/2.6.32-431.5.1.el6.x86_64/extra foonsd2.example.org: Unloading module mmfs26 foonsd1.example.org: 'shutdown' command about to kill process 3242 foonsd1.example.org: Unloading modules from /lib/modules/2.6.32-431.5.1.el6.x86_64/extra foonsd1.example.org: Unloading module mmfs26 foonsd2.example.org: Unloading module mmfslinux foonsd2.example.org: Unloading module tracedev foonsd1.example.org: Unloading module mmfslinux foonsd1.example.org: Unloading module tracedev Sun Mar 23 14:20:53 MST 2014: mmshutdown: Finished # mmchnode --client --noquorum -N foonsd1,foonsd2 Sun Mar 23 14:21:04 MST 2014: mmchnode: Processing node foonsd1.example.org Sun Mar 23 14:21:04 MST 2014: mmchnode: Processing node foonsd2.example.org Verifying GPFS is stopped on all affected nodes ... mmchnode: Propagating the cluster configuration data to all affected nodes. This is an asynchronous process. # mmlscluster GPFS cluster information ======================== GPFS cluster name: fooamcp.example.org GPFS cluster id: 10159024612008204251 GPFS UID domain: fooamcp.example.org Remote shell command: /usr/bin/ssh Remote file copy command: /usr/bin/scp GPFS cluster configuration servers: ----------------------------------- Primary server: foonsd1.example.org Secondary server: foonsd2.example.org Node Daemon node name IP address Admin node name Designation ------------------------------------------------------------------------------ 1 foo01.example.org 140.252.27.21 foo01.example.org 2 foo02.example.org 140.252.27.22 foo02.example.org 3 foo03.example.org 140.252.27.23 foo03.example.org 4 foo04.example.org 140.252.27.24 foo04.example.org 5 foo05.example.org 140.252.27.25 foo05.example.org 7 foo07.example.org 140.252.27.27 foo07.example.org 8 foo08.example.org 140.252.27.28 foo08.example.org 9 foo09.example.org 140.252.27.29 foo09.example.org 10 foo10.example.org 140.252.27.30 foo10.example.org 11 foo11.example.org 140.252.27.31 foo11.example.org 12 foo12.example.org 140.252.27.32 foo12.example.org 13 foo13.example.org 140.252.27.33 foo13.example.org 14 foo14.example.org 140.252.27.17 foo14.example.org 15 foo15.example.org 140.252.27.18 foo15.example.org 16 foonsd1.example.org 140.252.27.15 foonsd1.example.org 17 foonsd2.example.org 140.252.27.16 foonsd2.example.org 18 foo06.example.org 140.252.27.26 foo06.example.org 19 foodb1.example.org 140.252.27.14 foodb1.example.org 20 foonsd5.example.org 140.252.27.36 foonsd5.example.org quorum-manager 21 foonsd3.example.org 140.252.27.34 foonsd3.example.org quorum-manager 22 foonsd4.example.org 140.252.27.35 foonsd4.example.org quorum-manager
2014-04-13 | 2 Comments
It is possible to move a complete GPFS filesystem between clusters with the mmexportfs and mmimportfs commands. I’ve found this useful for preparing a complete filesystem and populating it with data before shipping it to a remote site.
Preparing the mss3 filesystem to be exported/relocated to the another GPFS cluster attaching to the GPFS NSD server named bar1.example.com:
# mmlsdisk mss3 disk driver sector failure holds holds storage name type size group metadata data status availability pool ------------ -------- ------ ------- -------- ----- ------------- ------------ ------------ bar1_nsd1 nsd 512 1 Yes Yes ready up system bar1_nsd2 nsd 512 1 Yes Yes ready up system bar1_nsd3 nsd 512 1 Yes Yes ready up system bar1_nsd4 nsd 512 1 Yes Yes ready up system # mmumount mss3 -a Wed May 29 12:49:56 MST 2013: mmumount: Unmounting file systems ... # mmexportfs mss3 -o mss3.gpfs mmexportfs: Processing file system mss3 ... mmexportfs: Propagating the cluster configuration data to all affected nodes. This is an asynchronous process. # scp mss3.gpfs bar1.example.com:. mss3.gpfs 100% 3379 3.3KB/s 00:00
2014-04-13 | 0 comments
cut is an incredibly use command but it can be frustrating to use on columar data with a variable number of spaces between columns. awk is often a better alternative in that case but, IMHO, it often feels more nature to use cut in a shell pipeline rather than a quoted onliner awk expression.
I recently stumbled across this trick (which I suspect I’ve discovered and forgotten more than once) of using tr to compress multi-space field seperators into a since character.
Consider this simple example of finding the ammount of free memory on the system:
$ free
total used free shared buffers cached
Mem: 7731688 5692336 2039352 396756 740020 2239136
-/+ buffers/cache: 2713180 5018508
Swap: 7733244 0 7733244
$ free | grep Mem: | cut -d ' ' -f4
$ free | grep Mem: | tr -s ' ' | cut -d ' ' -f4
2037816
2014-04-09 | 0 comments
(If you haven’t heard about the heartbleed vulnerability yet you should consider to coming out of your pocket universe more often…)
In an amazingly short turn around time since the issue was publicly disclosed, Filippo Valsorda has put up a site called http://filippo.io/Heartbleed/ that will check URLs for the heartbleed bug/vulnerability. This has turned out to be handy this morning for catching several sites that had patched openssl package installed last night but the services hadn’t yet been restarted. Kudos to Filippo!
The scanner service in action:
UPDATE:
The code for the site is on github https://github.com/FiloSottile/Heartbleed and is definitely worthy of being starred.