4

Currently have a client stating that the 5 users on a 2cpu, 4gb mem app server are experiencing slowness and often, locking up for moments at a clip.

Framework is a VM running in vSphere 5.5 [intel based host, was recently migrated from an AMD host], Server 2008 r2, Citrix 6.5, local machine using Receiver 4.2.

I've upped CPU, which only increased cpu ready and made perf worse. Increased memory to 8Gb, but the VM currently sits at a 3.8Gb mem used plateau throughtout the day, so that didnt help any.

Updated citrix to rollup 5, windows updates, poured through event logs and I have hit a wall as where to go.

Currently working on a performance monitor set that Ill turn into a data collect set and was wondering what would be "must have" counters included for a typical "citrix slowness" investigation. 

Counters:
Citrix MetaFrame Pesentation Server - Cumulative Server Load
ICA Session - Latency - Session average
Memory - Pages/Sec
Paging File - % Usage

What else would anyone suggest I place into this set that could leave me to anything that I could hang my hat on and move forward with?

Thanks.

soMuch2Learn
  • 333
  • 1
  • 6
  • 16
  • Locally on the VM you will want to especially monitor context switching in addition to the usual Windows counters, and depending on the individual application resource usage profiles you may need additional counters. But as you are running virtual you need to monitor the host as well, and contention between the virtual machines too. Adding vCPUs to a virtual machine can degrade performance, I suggest a light readthrough of https://www.vmware.com/files/pdf/techpaper/VMware-vSphere-CPU-Sched-Perf.pdf – ErikE Jul 30 '15 at 09:10
  • Thank you. Yes, i learned pretty quickly what CPU ready is when I thought CPU was the issue. Thank you, ill check out context switching and see if I can have the infra team look into the host. – soMuch2Learn Jul 31 '15 at 05:06
  • Citrix monitoring is tricky. Many a times, the issue may not be caused by the Citrix server itself. Hence, you may want to make sure you are monitoring your Citrix servers and also the infrastructure services that Citrix depends on. Consider a holistic monitoring solution for Citrix. There are several options. –  Aug 02 '15 at 02:19
  • We've looked at the host, which iniitally was AMD, but I migrated it to an Intel cluster that did increase performance by a bit, however, still having issues with apps locking up. Various apps, no commonality. @ErikE, I have been looking at the context switching and am looking at 27,000 switches/sec over 2 vCPUs, but trying to pin down how horrible this is. According to Process Explorer, hosted outlook is the culprit for high context switches, possibly due to a CRM plugin. – soMuch2Learn Aug 05 '15 at 18:23
  • Ok, so if possible a test where the plugin is disabled and a test with no outlook running at all for problem isolation? Note also that if you got the metric from within the VM it will be an indicator but not an accurate number due to time sync issues, accurate time based metrics can only be pulled from the host level. You could also do tests where you decrease vCPU to one (which may actually increase performance), and where you disable unused virtual hardware (to minimize hardware interrupt noise in the context switching metric). I would also recommend taking a look at VM contention. – ErikE Aug 05 '15 at 21:06
  • ...and contention for any shared underlying resources such as SAN etc. – ErikE Aug 05 '15 at 21:09

0 Answers0