[PVE-User] about IO Delay using openvz and zimbra
Fabrizio Cuseo
f.cuseo at panservice.it
Fri Jun 17 14:42:44 CEST 2016
Yes, shared storage is a must; you can (live) migrate VMs and better scale your cluster.
For you problem, try to debug (with dstat and sar) how much your disk is used (on both container and host); disk performances depends too much from disks, controllers, cache; and cheap sata disk with "dumb" controllers are not the best choice for a zimbra server.
I have several zimbra installations, with sata disks too, but using 6 sata disks, and memory cached raid-10 controllers.
PS: how many users do you have in your zimbra server ?
----- Il 17-giu-16, alle 14:25, Orlando Martinez Bao ombao at unah.edu.cu ha scritto:
> Hello, Havana is great.
>
> My cluster configuration is as follow.
>
> I have 7 node all use local storage the Servers descriptions is as follow:
> 1. Dell PowerEdge 1xCPU 8GRAM 3HDD 10K 600GB not in RAID.
> 2. Dell PowerEdge T320 1xCPU, 12GRAM, 2HDD 1Tera RAID-1 disk Hardware.
> (Server where I have a container with Zimbra)
> 3. Inspur 2xCPU (24 core), 64GRAM, 5HDD 1 Tera RAID-5 Hardware.
> All other server are PC I5 8GRAM and 1 HDD 1TB.
>
> I have 2 PC with 4xHDD 1TB, 1 with FreeNAS and 1 with Openfiler Both I use
> with NFS to backup and storage data.
>
> Al VM in the cluster use local storage, Why? Because never could create a VM
> using NFS shared storage. I did two week ago so only 1 VM is shared storage.
>
> I want to create a new cluster with Proxmox 4.2 using Ceph for storage, but
> for now I need to solve this problem.
>
> Thanks
> Best regards
> Orlando
>
>
>
> -----Mensaje original-----
> De: pve-user [mailto:pve-user-bounces at pve.proxmox.com] En nombre de Fabrizio
> Cuseo
> Enviado el: viernes, 17 de junio de 2016 4:36
> Para: pve-user
> Asunto: Re: [PVE-User] about IO Delay using openvz and zimbra
>
> Hello.
> First, please say "hello" to Havana... i miss Cuba (my only time in cuba, 2
> years ago, habana, trinidad, vinales).
>
> Regarding your problem: can you describe better your cluster configuration ?
> Which kind of storage (both local and shared) are you using ? RAID-1 disks
> are software or hardware raid ? Do you have a controller with cache memory
> and backup battery ?
>
> Why have you choosen to have only a container with local storage (and not a
> VM with shared storage) for zimbra installation ?
> If you have a very busy zimbra server, you will have a lot of disk I/O
> activity, and hardware requirements are growing for last zimbra releases
> (8.6). You can try to monitor your container activity with both "sar" and
> "dstat".
>
> Regards, Fabrizio
>
>
> ----- Il 17-giu-16, alle 4:38, Orlando Martinez Bao ombao at unah.edu.cu ha
> scritto:
>
>> Hello friends
>>
>> I am SysAdmin at the Agrarian University of Havana, Cuba.
>>
>>
>>
>> I have installed Proxmox v3.4 here a cluster of seven nodes and for
>> some days I am having problems with a node in the cluster which only
>> has a Container with Zimbra 8.
>>
>> The problem is I'm having a lot of I / O Delay and that server is very
>> slow to the point that sometimes the service is down.
>>
>> The server is PowerEdge T320 Dell with Intel Xeon E5-2420 12gram with
>> 12 cores and 1GB of disk 7200RPM 2xHDD are configured as RAID1.
>>
>> I have virtualizing the zimbra 8 using a template of 12.04, has 8
>> cores, 8G RAM, 500GHD the storage of container is local storage.
>>
>> Then I put the output to see the IO when you are not running the VM.
>> Look at the BUFFERED READS that are marked are very bad. And those
>> moments have seen IO Delay up to 50%.
>>
>>
>>
>> root at n07:~# pveperf
>>
>> CPU BOGOMIPS: 45601.20
>>
>> REGEX/SECOND: 1025079
>>
>> HD SIZE: 9.84 GB (/dev/mapper/pve-root)
>>
>> BUFFERED READS: 1.51 MB/sec
>>
>> AVERAGE SEEK TIME: 165.88 ms
>>
>> FSYNCS/SECOND: 0.40
>>
>> DNS EXT: 206.20 ms
>>
>> DNS INT: 0.91 ms (unah.edu.cu)
>>
>> root at n07:~# pveperf
>>
>> CPU BOGOMIPS: 45601.20
>>
>> REGEX/SECOND: 1048361
>>
>> HD SIZE: 9.84 GB (/dev/mapper/pve-root)
>>
>> BUFFERED READS: 0.78 MB/sec
>>
>> AVERAGE SEEK TIME: 283.84 ms
>>
>> FSYNCS/SECOND: 0.50
>>
>> DNS EXT: 206.13 ms
>>
>> DNS INT: 0.89 ms (unah.edu.cu)
>>
>> root at n07:~# pveperf (Este fue cuando detuve la VM)
>>
>> CPU BOGOMIPS: 45601.20
>>
>> REGEX/SECOND: 1073712
>>
>> HD SIZE: 9.84 GB (/dev/mapper/pve-root)
>>
>> BUFFERED READS: 113.04 MB/sec
>>
>> AVERAGE SEEK TIME: 13.49 ms
>>
>> FSYNCS/SECOND: 9.66
>>
>> DNS EXT: 198.59 ms
>>
>> DNS INT: 0.86 ms (unah.edu.cu)
>>
>> root at n07:~# pveperf
>>
>> CPU BOGOMIPS: 45601.20
>>
>> REGEX/SECOND: 1024213
>>
>> HD SIZE: 9.84 GB (/dev/mapper/pve-root)
>>
>> BUFFERED READS: 164.30 MB/sec
>>
>> AVERAGE SEEK TIME: 13.61 ms
>>
>> FSYNCS/SECOND: 16.34
>>
>> DNS EXT: 234.75 ms
>>
>> DNS INT: 0.94 ms (unah.edu.cu)
>>
>>
>>
>>
>>
>> Please help me.
>>
>> Best Regards
>>
>> Orlando
>>
>>
>>
>> _______________________________________________
>> pve-user mailing list
>> pve-user at pve.proxmox.com
>> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
>
> --
> ---
> Fabrizio Cuseo - mailto:f.cuseo at panservice.it Direzione Generale -
> Panservice InterNetWorking Servizi Professionali per Internet ed il
> Networking Panservice e' associata AIIP - RIPE Local Registry
> Phone: +39 0773 410020 - Fax: +39 0773 470219 http://www.panservice.it
> mailto:info at panservice.it Numero verde nazionale: 800 901492
> _______________________________________________
> pve-user mailing list
> pve-user at pve.proxmox.com
> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
--
---
Fabrizio Cuseo - mailto:f.cuseo at panservice.it
Direzione Generale - Panservice InterNetWorking
Servizi Professionali per Internet ed il Networking
Panservice e' associata AIIP - RIPE Local Registry
Phone: +39 0773 410020 - Fax: +39 0773 470219
http://www.panservice.it mailto:info at panservice.it
Numero verde nazionale: 800 901492
More information about the pve-user
mailing list