[OmniOS-discuss] Poor read performance on fresh zpool
Hi, on a OmniosCE r151022 system we have rebuild our zfs pool. The pool is constructed out of 18 mirror vdevs, each consisting of 2 1,8TB SAS drives. Since about half the disks are 512b and the other half are 512e (physical 4k) we opted to use 4k sectors using sd.conf for all devices; thus all vdevs report as ashift 12. During testing today we experienced extremely poor read performance. Doing sequential reads with dd of the pool results in a read rate of 300MB/s, benchmarking with bonnie++ gives about 600MB/s Write performance is as expected, around 1,4GB/s. Scrub also performs as expected, zpool status shows a scrub speed of 2GB/s, iostat -x shows a total pool speed of 4GB/s. Any hints as to what might be causing our poor read performance? Sincerely, Gijs Peskens ___ OmniOS-discuss mailing list OmniOS-discuss@lists.omniti.com http://lists.omniti.com/mailman/listinfo/omnios-discuss
Re: [OmniOS-discuss] VEEAM backups to CIFS fail only on OmniOS hypercongerged VM
> On Jun 14, 2018, at 3:04 AM, Oliver Weinmann wrote: > I would be more than happy to use a zone instead of a full blown VM but since > there is no ISCSI and NFS server support in a Zone I have to stick with the > VM as we need NFS since the VM is also a datastore for a few VMs. You rambled a bit here, so I'm not sure what exactly you're asking. I do know that: - CIFS-server-in-a-zone should work - NFS-server-in-a-zone and iSCSI-target-in-a-zone are both not available right now. There is purported to be a prototype of NFS-server-in-a-zone kicking around *somewhere* but that may have been tied up. I'd watch distros, especially those working on file service, to see if that shows up at some point, where it can be upstreamed to illumos-gate (and then back down to illumos-omnios). Dan ___ OmniOS-discuss mailing list OmniOS-discuss@lists.omniti.com http://lists.omniti.com/mailman/listinfo/omnios-discuss
[OmniOS-discuss] VEEAM backups to CIFS fail only on OmniOS hypercongerged VM
Dear All, I’m struggling with this issue since day one and I have not found any solution for it yet. We use VEEAM to back up our VMs and an OmniOS VM as a CIFS target. We have one OmniOS VM for internal and one for DMZ. VEEAM Backups to the one for internal work fine. No problems at all. Backups to the DMZ one fail every time. I can access the CIFS share just fine from windows. When the backup starts two or three VMs are backed up and then it fails. I have requested support from VEEAM and it turns out the same job running against a Windows server CIFS share works just fine. I couldn’t believe that OmniOS is the culprit as the CIFS implementation from Illumos is very good. So I setup a new OmniOS bare-metal server and created a zone for DMZ. I setup a cifs share and ran the same job. Everything works fine. I compared the settings from both the VM and the Zone and they are 100% identical. Only difference is one is a VM and one is a Zone. But since the VEEAM backup to the internal VM has no problems with the backup, I don’t think virtualization is a problem here. Is there anywhere I can start investigating further? I would be more than happy to use a zone instead of a full blown VM but since there is no ISCSI and NFS server support in a Zone I have to stick with the VM as we need NFS since the VM is also a datastore for a few VMs. Any help is really appreciated. Best Regards, Oliver___ OmniOS-discuss mailing list OmniOS-discuss@lists.omniti.com http://lists.omniti.com/mailman/listinfo/omnios-discuss