• State Not Answered
  • Locked Locked
  • Replies 5 replies
  • Subscribers 27 subscribers
  • Views 6523 views
  • Users 0 members are here
Options
Related
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Server Performance Impact when RR is exporting volumes to the core

ncates
over 7 years ago

We have two Windows 2012 servers running HyperV guests. The hosts are each running three guests. We are experiencing significant performance hits on the guest machines on one of the hosts, when RR is trying to export the volumes to the core. The other host/guests do not have any problems. We have been working with both Quest and Dell support to try and resolve this - so far no luck.

What we have three Windows 2012 R2 Hyper-V servers running on Host1. These guest servers are all protected by RR. One of the guests is a DC, one is a document imaging system, the other is a terminal server (remote desktop services). When the backup snapshot and subsequent export of data, to the RR core happens on one of these guests, we experience a performance hit on the host and all of the guest servers. The guest become practically unusable and we typically have to stop the export job and stop the service on the guest in order to get things to return to a usable condition. The export of the volumes is very slow when this happens. This behavior does not happen with the other host server or the guests on that host, that are also protected by RR. If we look at the resource monitor in Windows and go to the disk tab - we see high disk queue length activity on the affected host/guests. This symptom does not manifest itself on the other host/guests. They all have the same agent version of RR. At this point we are not sure what this means. We do not know if RR is at fault or if we have a hardware or configuration problem of some sort on host 1.

Has anyone else experienced this?

  • 0 over 7 years ago
    Hi ncates:
    Is the Core a physical or a virtual machine? If it is a physical machine is it the Hyper-V host Host1?
    We have some MSP customers that use Hyper-V hosts as cores and protect and export the VMs on the same host. In this case, an expected loss of performance is encountered when stand-by export operations are performed. It is not clear from your post where the exports are performed. Are they targeted to the same Hyper-V host on which the VMs are located? If it is the case, would it be possible to direct the exports to a different Hyper-V host in order to compare performance?
    Hope that this helps.
  • 0 over 7 years ago
    Hello, and thanks for the reply. The core is a physical machine. It is a Dell DL4300 appliance - dedicated as RR core.The two host machines, with the guest servers that are protected are on two separate physical servers. So we have the following:
    A) Dell DL4300 RR Core at site1
    B) Dell T620 Windows 2012 R2 Hyper-V host1 - with three guest servers - protected by RR - backups going to the DL4300 at site 1
    C) Dell T630 Windows 2012 R2 Hyper-V host2 - with three guest servers - protected by the RR - backups going to the DL4300 at site 1
    D) Another Dell physical server serving as a replication core - at site 2 (offsite)

    We are having the performance problems with the guests on host1 (B above). We are unsure if the problem is a RAID or disk problem on the host1 server - or a RR software problem.
  • 0 over 7 years ago
    I notice that the server with the issue is a Gen 12 server while the one that works is a Gen 13.
    One of the things to do is to check the drivers and firmware (storage, networking etc.) and update them. If time permits, the LifeCycle controller is the way to go. Additionally, making sure that the Raid Array caching is set up to write back/adaptive read may help.
    The same should be done for the DL (starting with applying the newest RUU 3.1.38 and adding RR 6.1.1.137 on top of it -- support.quest.com/.../download-new-releases). Oh -- if the Storage Controller Cache battery is dead, the controller will auto-configure to write-through to avoid data loss.
    RR 6.1.1.137 allows protecting hyper-V VMs agentlessly. If the protected machines are not Exchange/SQL application servers, it is worth trying it. If they are you need to see if any application related operations (mountability, attachability etc.)are important for you and decide accordingly.
    At last but not at least, I am not sure where you are doing the exports. If you do them on the DL4300 machine, I would pause them for a while and check the performance of the appliance. If the performance improves considerably, we need to check the storage configuration on the DL and update it as needed.
    Please let us know.
  • 0 over 7 years ago
    I seem to remember a doc that had some best practices for clients that are VM's. Most of them were ESX I think but not all. But I have been looking for this and not finding it.

    You may want to look at the VM settings between the clients that are having this issue and ones that are not. Are they on different disk arrays, different hardware settings (and drivers like Tudor mentioned) Check out some performance tuning docs from the Hyper-V manufacture.

    I think one of the points was to make sure you have VM tools installed. I know you want this for agentless but I think it was a performance recommendation also

    Tudor, do you know of any recommendations specific to VM clients? (ESXi or HyperV)
  • 0 over 7 years ago
    Thanks for the replies. We are going to work through the list an see where we are at in a few days. Will post back with the results