> Andrew,
> I'm not sure who says that your data will be corrupted, that's one
> complaint I haven't heard. IMO the problems are mostly related to
> operational behavior, failure modes and resource contention.
> By running over NFS you introduce latency into your IO; MySQL can be
> particularly sensitive to such things. Now, if your network may be very
> fast and low latency and everything will work great, but now what happens
> when someone else in your environment starts doing a massive backup and
> fills up bandwidth on a link that your NFS traffic must use? The result
> will be degraded performance for your databases.
> The NFS device itself becomes a new thing to fail. If it does, then all
> databases which use it will be unavailable. You can try to make this
> device more reliable in many ways, but now your operations depends on how
> well you understand this other system.
> Also, if you have many instances running on on NFS device, what happens
> when your production environment starts sending a lot more queries or a
> really IO intensive pattern of queries? They compete for resources now; if
> one does a lot of IO, it can slow down others, which makes your load issue
> worse at an innopportune moment. You wind up with the potential for
> cascading type failures.
> -Gavin Towey
> On Thu, Nov 8, 2012 at 2:23 AM, Andrew Holway <andrew.hol...@gmail.com>wrote:
>> I have a fast ZFS/NFS box. Its very, very fast and very silky.
>> I've been running KVM machines on this NFS, 48 VMs on 4 servers. The
>> servers are Dual socket boxes (Intel E5-2670) with 144GB ram each.
>> I have been benchmarking the whole platform with tpcc-mysql and, when
>> just performing read testing (number of warehouses much smaller then my
>> InnoDB cache size), I saturate the CPU before the ZFS/NFS box.
>> Some people have suggested that running Mysql on NFS is a terrible idea.
>> That my database will be irreparably corrupted. They mutter something about
>> locking and terrible performance.
>> Thing is, the performance is really really great and it seems quite
>> stable. Can someone tell me why this is a bad idea?
>> Thanks,
>> Andrew
>> --
>> You received this message because you are subscribed to the Google Groups
>> "Percona Discussion" group.
>> To post to this group, send email to percona-discussion@googlegroups.com.
> --
> You received this message because you are subscribed to the Google Groups
> "Percona Discussion" group.
> To post to this group, send email to percona-discussion@googlegroups.com.
