Use api to extract temperature data?

[ROWENA_2]

Hi fellows,

I have a HP Proliant servers gathered in a server room not very well cooled. The airco once failed and a server died. It gave me the urge to start monitoring the temperature and humidity of those servers.

1st: i would like to know what is a critical temperature for a server
2nd: i would like to know what i can do to monitor the temperature of the servers.

I would like to use PHP language to use API to extract my server's temperature and humidity data. Is there possible to do so?

If you have any help to provide, i would like to thank you in advance!
All the best.

[Matt Palmer_2 · Accepted solution]

Hi,

which OS are the servers runnning. Is it the same on all servers?

In linux you can use "ipmitool sensor ",eg "ipmitool sensor list".

this will output all of the temperature info

check this link:

http://forums13.itrc.hp.com/service/forums/questionanswer.do?threadId=1353435

which discusses a windows version of the tool.


hope that helps

regards

Matt

[ROWENA_2]

Server is running in XENCenter OS

[Matt Palmer_2]

Hi,

look here for Xenserver and IPMI

http://forums.citrix.com/message.jspa?messageID=1393121

regards

Matt

[David Claypool]

@ROWENA, ProLiant servers have intelligent management that monitors thermal conditions and matches them against individually-calibrated caution and critical thresholds and can generate alerts when those are crossed, allowing you to potentially rectify the situation if possible. A ProLiant G6 server has up to 32 of these sensors depending on the platform. Additionally, this intelligence will initiate a graceful shutdown when it reaches the critical state to avoid damage to the hardware.

Since you report that you had servers that "died," my guess is that you haven't availed yourself of the ProLiant management agents delivered with the ProLiant Support Pack (either SNMP or WBEM), or at least you hadn't configured them to send an alert (SNMP trap, WBEM indication or SMTP mail; you can even trigger a script to run). I further am guessing that you have disabled the 'Automated Server Recovery' feature that would have initiated a graceful shutdown, protecting the hardware and avoiding data loss, when the server crossed the critical threshold.

Based on the above, it sounds like your solution is to query the sensors at some frequency through a script. That sounds a little 1980s to me. First, your script is going to be fairly complicated (you'll need to check anywhere from 1 to n sensors depending on the platform involved) and don't forget that the data is not meaningful unless you compare that value to the threshold for the thermal zone. Even if you set this up in a virtual machine, you're going to have a tremendous investment in time to create and maintain this script, not to mention the waste of bandwidth and CPU resources especially considering the system is intelligent enough to monitor itself constantly and you get this for free.

As an added bonus you will get pre-failure notifications on CPU, ECC memory and disk errors, which is invaluable in a virtualized environment because you have more eggs in each basket. In total there are more than 1500 items monitored by the ProLiant management agents and this is designed to maximize system availability by avoiding unplanned downtime.

We additionally provide a tool, HP Systems Insight Manager, to act as a console to aggregate the information from multiple systems --> http://www.hp.com/go/hpsim but if you are perhaps using Citrix Xen Essentials for that role, we have provided integration to that console so it can benefit from the intelligence of ProLiant server management --> http://www.hp.com/go/citrix