Hi,
have here a 2 host, same configuration, one with a 10Gbit iSCSI attached Exomium box and one with a 10Gbit iSCSI QNAP TS-879 box (direct attached).
Have installed 6 Hitachi 2TB HDD in the QNAP. Did a lot of testing with QNAP in advance to move the box in production (FW:3.7.1).
Both installations ran fine and the performance was very well for a couple of month.
This is still true for the Exomium attached host, but after a firmware upgrade of the QNAP box (because a important fix came with this version),
the attached Vsphere 5.1 host freezed now 3 times in 3 weeks. The datastore disk access latency is bad, compared to the other box.
I ran "iostat -x 5" on the QNAP box, the HDD busy% is less than 30%, service time 3-5 ms, this seams normal.
Here are some log, no message on the Hardware console, according to the SNMP manager the Vsphere host stopped 11/23/12 21:02:05.
1.) Any idea to proof, if this performance behavior causes the frezze? Any way to do a better diagnose?
2.) Anybody else with Qnaps and Vsphere 5.1? Have ran "iostat", "top" etc., checked the logs, anything else to get more informations?
3.) Have still a free 10Gbit port on both sides, anybody with multipathing on Qnaps?
Thanks
Henri
2012-10-23T20:55:01Z crond[4569]: crond: USER root pid 901826 cmd /sbin/hostd-probe
2012-10-23T20:55:01Z syslog[901827]: starting hostd probing.
2012-10-23T20:55:16Z syslog[901827]: hostd probing is done.
2012-10-23T21:00:01Z crond[4569]: crond: USER root pid 902032 cmd /usr/lib/vmware/vmksummary/log-heartbeat.py
2012-10-23T21:00:01Z crond[4569]: crond: USER root pid 902033 cmd /sbin/hostd-probe
2012-10-23T21:00:01Z syslog[902035]: starting hostd probing.
2012-10-23T21:00:16Z syslog[902035]: hostd probing is done.
2012-10-23T21:01:01Z crond[4569]: crond: USER root pid 902117 cmd /sbin/auto-backup.sh
2012-10-24T05:57:12Z watchdog-vobd: [4514] Begin '/usr/lib/vmware/vob/bin/vobd ++min=0,max=100,group=uwdaemons', min-uptime = 60, max-quick-failures = 1, max-total-failures = 1000000, bg_pid_file = ''
2012-10-24T05:57:12Z watchdog-vobd: Executing '/usr/lib/vmware/vob/bin/vobd ++min=0,max=100,group=host/vim/vmvisor/uwdaemons'
2012-10-24T05:57:14Z jumpstart: Using policy dir /etc/vmware/secpolicy
2012-10-24T05:57:14Z jumpstart: Parsed all objects
2012-10-24T05:57:14Z jumpstart: Objects defined and obsolete objects removed
2012-10-24T05:57:14Z jumpstart: Parsed all domain names
2012-10-23T21:01:40.070Z [2B09BB90 verbose 'vim.PerformanceManager'] HostCtl Exception in stats collection: Sysinfo error on operation returned status : Not found. Please see the VMkernel log for detailed error information
2012-10-23T21:01:40.070Z [2B09BB90 verbose 'vim.PerformanceManager'] HostCtl Exception in stats collection. Turn on 'trivia' log for details
2012-10-23T21:02:00.026Z [2B05AB90 warning 'vim.PerformanceManager'] Calculated read I/O size 602112 for scsi0:0 is out of range -- 602112,prevBytes = 1577965419008 curBytes = 1577995524608 prevCommands = 56798322curCommands = 56798372
2012-10-23T21:02:00.063Z [2B05AB90 verbose 'vim.PerformanceManager'] HostCtl Exception in stats collection: Sysinfo error on operation returned status : Not found. Please see the VMkernel log for detailed error information
2012-10-23T21:02:00.063Z [2B05AB90 verbose 'vim.PerformanceManager'] HostCtl Exception in stats collection. Turn on 'trivia' log for details
2012-10-23T21:02:03.585Z [2AD81B90 verbose 'SoapAdapter'] Responded to service state request
2012-10-23T21:02:04.477Z [2A981B90 verbose 'Cimsvc'] Ticket issued for CIMOM version 1.0, user root
2012-10-23T21:02:19.680Z [2AD81B90 verbose 'DvsManager'] PersistAllDvsInfo called
2012-10-23T21:02:20.023Z [2B019B90 warning 'vim.PerformanceManager'] Calculated read I/O size 577308 for scsi0:0 is out of range -- 577308,prevBytes = 1577995524608 curBytes = 1578026699264 prevCommands = 56798372curCommands = 56798426
2012-10-23T21:02:20.057Z [2B019B90 verbose 'vim.PerformanceManager'] HostCtl Exception in stats collection: Sysinfo error on operation returned status : Not found. Please see the VMkernel log for detailed error information
2012-10-23T21:02:20.057Z [2B019B90 verbose 'vim.PerformanceManager'] HostCtl Exception in stats collection. Turn on 'trivia' log for details
2012-10-23T21:02:24.707Z [2B09BB90 info 'Snmpsvc'] UpdateStats: report cimom converter stats started
2012-10-23T21:02:24.707Z [2B09BB90 info 'Snmpsvc'] DumpStats: cimom stats file /tmp/.cvtcimsnmp.xml generated, size 298 bytes.
2012-10-23T21:02:24.707Z [2B09BB90 info 'Snmpsvc'] PublishReport: file /tmp/.cvtcimsnmp.xml published as /tmp/cvtcimsnmp.xml
2012-10-23T21:02:24.707Z [2B09BB90 info 'Snmpsvc'] DumpStats: cimom stats file published
2012-10-23T21:02:24.707Z [2B09BB90 info 'Snmpsvc'] NotifyAgent: write(51, /var/run/snmp.ctl, N) 1 bytes to snmpd
2012-10-23T21:02:24.707Z [2B09BB90 info 'Snmpsvc'] UpdateStats: report cimom converter stats completed
Section for VMware ESX, pid=5084, version=5.1.0, build=799733, option=Release
------ In-memory logs start --------
mem> 2012-10-24T06:00:56.807Z [FFEA8D20 verbose 'Default'] No update forwarding configured
mem> 2012-10-24T06:00:56.807Z [FFEA8D20 info 'Default'] Supported VMs 87
mem> 2012-10-24T06:00:56.807Z [FFEA8D20 info 'Handle checker'] Setting system limit of 2222
mem> 2012-10-24T06:00:56.807Z [FFEA8D20 info 'Handle checker'] Set system limit to 2222
mem> 2012-10-24T06:00:56.807Z [FFEA8D20 info 'Default'] Setting malloc mmap threshold to 32 k
mem> 2012-10-24T06:00:56.807Z [FFEA8D20 info 'Default'] getrlimit(RLIMIT_NPROC): curr=64 max=128, return code = Success
mem> 2012-10-24T06:00:56.807Z [FFEA8D20 info 'Default'] setrlimit(RLIMIT_NPROC): curr=128 max=128, return code = Success
I've done quite a bit of work with the qnap, i'm currently running a EC1279U-RP and I gave up on iscsi. I found NFS to work much much better.
You should do very very well with nfs and an actual 10 gig link.
Hi,
have downgraded the firmware and run now 50% of the VMs on a NFS datastore. First seams to help, but as soon I have more I/O activity (20-30 MBytes/sec.) it's as worst as iSCSI. The migrate of a 300 GB VM from the iSCSI to the NFS datastore took a very long time and the disk latency got super bad (up to 500ms).
Any body else?
Thanks
Henri
I'm running firmware Version 3.7.1 build 20120614 if that helps.
Also what Raid configuration are you running on your box?
Hi,
I have here a Raid6 with 6 Hitachi 2TB drives.
Firmware is now 3.7.2..
Could you post a Latency diagram from vCenter while running some I/Os?
Thanks
Henri
try switching to raid 5, you will see much better results.
Hi,
Freeze was mainboard related, the memory controller of Slot 1-4 is defect. The server runs currently with slot 5-8 without further issues.
The performance problem still exists.
Is anybody out there, whos runs successfully vSphere 5.1 with the QNAP TS-879 or better and usualy HDDs?
Have installed 2 SSD in addition where the VMDKs lifes, this works fine, the performance is good. As soon I move some VM to
the RAID 6 (6 Hitachi 2TB drives) the performance is in avg. 120 ms, up to 600 ms per I/O. Had a lot of trouble with VEEAM at
this performance.
Thanks
Henri
I have major issues with QNAP nfs performance and ESX 5i and ESX 4i
TS-459U+
TS-469U-RP
TS-859U
All patched up to the latest QNAP firmware. direct attached to a 10G switch. the QNAPs are connected to 1G ports. I use the advance load balance. loaded with server HD's
What I have found is low IO's so running a few VM's 5-10 there are okay. I have tried iscsi and NFS .. BUT ( i found this out when i moved my VDP vm onto the QNAP ) when you have a lot of IOPS (also noticed when i moved my zabbix monitoring box onto the QNAP DS) the QNAP fail badly I start to get 500ms -1.5s delays. I also notice the throughout maxs at around 9-10MB/s ....
I have raised SR's with VMWare of cause they blame QNAP. I have raised 2 or 3 support request with QNAP and tried their forums. Best I got was join us on skype and we can diagnose problem ... yeah right, we tried a few things but they don't seem to have an idea and basically don't really care..
Now I am stuck with 3 QNAP device that basically I pieces of shit.... BUT ! I did some experimenting and found if I mount the space directly in the VM, ie nfs mount inside a linux vm or iscsi mount space directly into Windows i actually get some performance, alot better than what a iscsi or nfs backed VMWare Datastore might give me.
I also found a problem when the raid array rebuilds... performance is a dog, rebuild rate is around 8-10Mb/s ... for a 10Tb array that takes forever ... My home built system does 110Mb/s easy and still server multimedia. The response I got from the QNAP support .. OH wait till its finished and then we can look for any faults .... Not sure how you diagnose a problem once its finished ... But it leads me to think they have gone ultra cheap on the SATA/SAS bus
SO I WOULD STAY AWAY FROM QNAP IT YOUR PLAN TO DO ANY THING SERIOUS..
hi alex,
im interesting from your posting about load balancing. what i understand is you run 3 qnaps and use load balancing to backup vms using nfs, cmiiw. if that correct, is it load balancing for backup process? how you do to achieve that?