Unsolved
This post is more than 5 years old
12 Posts
0
1327
May 11th, 2015 05:00
Lun response time alerting
I have a server that is running Solaris that has very stringent response times requirements. The storage for this server is on a VMAX. The application has come to me to see if there is a way to alert when the response time for one or more of their Luns reaches a certain threshold. We monitor our storage using Unisphere, and there does not appear to be a way to alert on a single lun or a single storage group within Unisphere. It appears to be all or nothing. I am looking for ways to make this happen. Any help would be appreciated.
No Events found!
PedalHarder
3 Apprentice
•
465 Posts
0
May 11th, 2015 15:00
Lun based performance alerting is available in the ViPER SRM product.
Have you considered a DIY alert? Solaris iostat disk metrics contain service time (asvc_t)... a little scripting around the iostat command could get you what you are after.
sullysam1
12 Posts
0
May 12th, 2015 06:00
Jasonc,
Thanks for the response…I was trying to stay away from creating and running any adhoc scripts. I believe we are in the early stages of implementing VIPER SRM. I will check that out…
matthewvalentin
10 Posts
0
May 12th, 2015 20:00
Sullysam,
You can monitor response time at the storage group level within Unisphere. These metrics are gathered every 5 minutes and you can set a threshold of the response time and how often it exceeds during the polling interval. Like Jason stated, this can also be done within SRM but those alerts are based off of the polling interval set between the collector and Unisphere which is not as granular. I use these more for tracking historical alerts to determine performance trends. Scripting from the host with IOstat will work but would be from the hosts perspective which includes application issues, compute bottlenecks, fabric, etc. If looking from the array a scripted SYMSAT would work by adding the luns to a device group but could start causing some performance issues from your solutions enabler instance depending on how granular you get ( I believe you can go down to every 5 seconds but that's extremely aggressive). Lots of options so I hope any of these help.
sullysam1
12 Posts
0
May 13th, 2015 05:00
Thanks for the input! I would love to be able to do this by Storage Group, however, the threshold is set for ALL storage Groups. As you know, all storage groups are not created equal, and have different performance profiles. I am going to look into the device group scenario….
sullysam1
12 Posts
1
May 13th, 2015 07:00
Great! Thanks for the information! This will certainly help.
matthewvalentin
10 Posts
0
May 13th, 2015 07:00
Actually you can set a different threshold for specific storage groups.
sullysam1
12 Posts
0
May 20th, 2015 09:00
Now that I have this alert setup, I am unsure whether this alert will be sent to the SNMP server. It seems from the documenation that I have read that only Array Type alerts are sent to SNMP server through Unisphere. Is this correct? It did not appear that a Performance Type alert would be sent to the SNMP server.