Host-based backup of VMware vSphere VMs.
Post Reply
csmall
Enthusiast
Posts: 30
Liked: 2 times
Joined: Mar 20, 2018 1:19 pm
Full Name: Colin Small
Contact:

Linux Server NFS Repository & Performance

Post by csmall »

What are the optimal NFS mount settings for fstab? I'm using defaults on CentOS.

Also, how can I test to make sure I am getting optimal speeds/performance to the NFS export?
nitramd
Veteran
Posts: 297
Liked: 85 times
Joined: Feb 16, 2017 8:05 pm
Contact:

Re: Linux Server NFS Repository & Performance

Post by nitramd »

Colin,

My initial thought is to pump multiple TBs onto your Linux repo then look at the stats within VBR.

But, performance is much more than that, i.e. speed of network, disk type and speed thereof, memory, CPUs, etc.

Can you shed some light on your hardware setup?
csmall
Enthusiast
Posts: 30
Liked: 2 times
Joined: Mar 20, 2018 1:19 pm
Full Name: Colin Small
Contact:

Re: Linux Server NFS Repository & Performance

Post by csmall »

2 x Synology Rackstation (12 disks, RAID6, btrfs, 80TB, 10Gb) 1 for main backups and 1 for backup copies/archival, Veeam server and proxies are running in vSphere as vm's on a NetApp (this is also where the main backups come from).

I modified the Synology NFS settings to be 32k instead of 8k and turned off NFSv4 support. (this seems to have improved performance with NFS)

I'm using a Linux server vm with an NFS mount for the repository.

My issue is with backup copy jobs, merges and GFS. I feel like the normal copy job process has good speeds with a single job running it Veeam console will rport like 285MBps-3xxMBps (NFS Linux mount server reports interfaces hitting between 2.8-6Gbps on the job... but when it begins the merge process or the GFS merge process it seems to slow down A LOT. The synology shows minimal resource usage during the merge and GFS processes. Volume utilization is really high but disk low and network is really low.

The proxy servers don't show any high resource utilization.

What is in play when a merge is happening? Network at all? Is the majority of if only on the storage itself? The console doesn't give you any performance information on the merging process.

Case#: 03379108 (started this case because a GFS on a copy job will not move past 4% but it used to finish after a long time of running. Now it doesn't move.)
nitramd
Veteran
Posts: 297
Liked: 85 times
Joined: Feb 16, 2017 8:05 pm
Contact:

Re: Linux Server NFS Repository & Performance

Post by nitramd »

It appears that the slowdown is occurring on the RAID array. As you may know, RAID 6 is very, very fast for reads but very slow for writes.

If your repository has a RAID card with cache memory then this will help.

Can you post what Support suggests?
csmall
Enthusiast
Posts: 30
Liked: 2 times
Joined: Mar 20, 2018 1:19 pm
Full Name: Colin Small
Contact:

Re: Linux Server NFS Repository & Performance

Post by csmall »

I tend to agree with you there however, the same jobs were processing up until recently. Slowly.. but reasonable.

I could move the storage appliance to another RAID type for faster writes but at the cost of integrity etc..(not to mention I would have to erase all the existing backup files.)

1 Particular job will not move past 4% but Veeam support analyzed the logs and determined that it was not dead but just incredibly slow.

The writes seem to be OK with the regular copy job processes.. it is just the merge/GFS operations that seem to have trouble.
Post Reply

Who is online

Users browsing this forum: No registered users and 99 guests