GPFS Monitor suite installation how to
A suite of tools for monitoring GPFS clusters.
Status: Abandoned
Brought to you by:
cruff
Menu
▾
▴
I would like to install the GPFS monitor suite to monitor our site. However, the info on the website seems to be unclear. I just wondering if there is any document for me to install/configure the suite? I will appreciate for the help, thanks...
I've added an INSTALLATION-HINTS file with the outline of what is needed to be done.
Hi, Craig:
thanks for your reply few monthes ago and I appreciate a lot. I was busy for my TSM and other tasks so didn't reply soon. I've installed postgresql 9.4 and ganglia 3.7 on my gpfs monitor host. I also build gpfsmonitor database with gpfsmonitor account in postgresql.
Now I'm working on the gpfsmond configuration file on client end and would like to have gpfsmond to report expelling, long waiters logs:
1. I just wondering for the file gpfsmond configuration just take your sample in the hint and just modify $monitorHost, filesys_a,filesys_b then place in /etc/gpfsmond.d/?
2. in the gpfsmonitor-2.2 tar zip file, for ganglia client which is gpfs nodes, I think I need to place files in gpfsmonitor-2.2/sbin,lib,etc to the system path in /sbin,/lib,/etc right?
3. I just wondering the extensions in gpfsmonitor-2.2/etc/gpfsmond.d/ how I can use them? for example, the gpfsgood, to report to ganglia server? on ganglia client, how to configure to report gpfs to ganglia server?
Yes
I'm not sure I understand this question fully. Here is my guess as to answers that may clarify things.
A. A GPFS client node does not need to be a ganglia client itself. The only thing that needs to be installed on the GPFS client nodes is gpfsmond and the contents of /etc/gpfsmond.d.
B. The information gathered by gpfsmond is forwarded to the node, where the gpfsmonitor and gpfsreporter daemons are running, which does need to be able to send metrics to ganglia (see answers to 3 below). On this node you need to install the contents of the GPFS monitor suite tar file in an appropriate location.
3.
A. You just select the extensions you want and copy them into /etc/gpfsmond.d on the nodes on which gpfsmond is running. They are dynamically loaded when gpfsmond is started.
B. To send metrics to a ganglia instance, define "$ganglia = 1;" in /etc/gpfsreporter.conf, define any pseudo-IP address mappings for the cluster names (being sure to also place these definitions /etc/hosts also or ganglia will be confused or refuse to display the metrics). You also need to have the Ganglia::Gmetric::XS perl module installed. You need to create a suitable ganglia configuration file and set the variable $gmondConf variable in gpfsreporter to point to that file (which can be done in the /etc/gpfsreporter.conf file). Then start gpfsreporter.
Hi, Craig:
I still have problem on db. How I can setup the db fro ganglia to feed in the data in? Do I need to add tables or fields?