Quantcast
Channel: VMware Communities: Message List
Viewing all articles
Browse latest Browse all 251495

Using ESXtop to find smoking gun

$
0
0

I have an Oracle environment located in two separate data centers. In data center A all hardware (servers, network, storage) is managed by us.  We get very good performance there.

 

Data center B is cloud "ish".  We manage the servers but network and storage is provided by the data center.  In that data center we get extremely inconsistent performance.

 

What I mean by inconsistent is sometimes jobs in datacenter B run in 30 seconds, 2 minutes, or 30 minutes.  Datacenter A provides very consistent job times of under 30 minutes.  Both VMs are the same size and jobs are running up against the same data.  Host are both G7 HP servers running the same build of ESXi 5.1.  With limited vision in datacenter B troubleshooting is a little more difficult.  My support calls into them have only produced the typical everything looks good responses.

 

My gut tells me the storage of the cloud provider is causing the issue but I try not to do assumptions in this business.  My plan to track down the problem is to run ESXTOP over a 24 hour span on a host in both datacenters.  When the Oracle team reports a slow job I want to compare the stats between the two hosts during that time frame in order to identify what metric might be causing the slowdown.  Armed with that data I can then work towards the resolution.

 

Here is the reason for my post as I see a couple of problems.

 

1.)   Will running ESXTOP for 24 hours even be possible?  This will likely produce a huge file.  So will it even be viewable?

2.)    Running ESXTOP in shorter spans has produced so much data it’s a bit overwhelming to view from an approach of I don’t know what’s wrong exactly so I want to look at everything.  How would you recommend I approach mining this data?


Viewing all articles
Browse latest Browse all 251495

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>