Re: VMWare file system / database corruption

Поиск
Список
Период
Сортировка
От Scot Kreienkamp
Тема Re: VMWare file system / database corruption
Дата
Msg-id 6013F25F-C4AB-4512-A9D0-B05DDB0953ED@la-z-boy.com
обсуждение исходный текст
Ответ на Re: VMWare file system / database corruption  (John R Pierce <pierce@hogranch.com>)
Список pgsql-general
We have fiber channel in production and iscsi in development. Both work well but iscsi is definitely slower. In our case the iscsi disks are slower too so that likely affects our speed on iscsi. 

Sent from my iPod

On Sep 21, 2009, at 4:24 PM, "John R Pierce" <pierce@hogranch.com> wrote:

Scot Kreienkamp wrote:
> On the contrary, we've been running PG in production for years now under VMWare.  Same with MSSQL.  We've never had any problems.  Less so than an actual physical machine actually since we can move the server to different physical hardware on demand.  Also makes disaster recovery MUCH easier.
>
> However, VMWare does have its places.  A high usage database is not one of them, IMHO.  A moderately or less used one, depending on requirements and the hardware backing it, is often a good fit.  And I agree with Scott about the snapshots.  They do tend to cause temporary communication issues with a running virtual machine occasionally, regardless of OS or DB type.  (The benefits outweigh the risks 99% of the time though, with backups being that 1%.)  In my experience the level of interference from snapshotting a virtual machine also depends on the type and speed of your physical disks backing the VMWare host and the size of the virtual machine and any existing snapshot.  I've been told that in VSPhere (VMWare 4.0) this will be significantly improved. 
>  

does your VMWARE server use NFS to communicate with the disks?   It was
my understanding most folks used SAN logical units for the virtual disks
with VMware ESX, and not NFS/NAS






В списке pgsql-general по дате отправления:

Предыдущее
От: John R Pierce
Дата:
Сообщение: Re: VMWare file system / database corruption
Следующее
От: Mark Felegyhazi
Дата:
Сообщение: Insert unique fails, still increments ID in a lookup table