<div dir="ltr">Hi Paul,<div><br></div><div>Thanks for the updates...</div><div><span style="font-size:12.8000001907349px"><br></span></div><div><span style="font-size:12.8000001907349px">A few things to clarify</span><br style="font-size:12.8000001907349px"><span style="font-size:12.8000001907349px">1) are all 4 for Gluster or are they the whole cluster.</span><span style="font-size:12.8000001907349px">keep in mind that Ovirt expects that you are running Gluster on </span><span style="font-size:12.8000001907349px">different nodes than your VM's. while I do not completely agree with </span><span style="font-size:12.8000001907349px">that idea that is how it is now.</span></div><div>Comment :- All 4 servers are the part of the one gluster volume...<br style="font-size:12.8000001907349px"><br style="font-size:12.8000001907349px"><span style="font-size:12.8000001907349px">2) RAID hardware is not necessarily a bad thing, in fact the cache may </span><span style="font-size:12.8000001907349px">help significantly. that said it is not required.</span><span style="font-size:12.8000001907349px">things to keep in mind if you are using RAID hardware for performance..</span><br style="font-size:12.8000001907349px"><span style="font-size:12.8000001907349px">2.a) RAID 0 will give you the best speed if you are using 3 nodes for </span><span style="font-size:12.8000001907349px">quorum. keep in mind this will require more work in the case of a </span><span style="font-size:12.8000001907349px">drive failure to remake the brick.</span><br style="font-size:12.8000001907349px"><span style="font-size:12.8000001907349px">2.b) if you are only doing a mirror across 2 bricks RAID 0 may be </span><span style="font-size:12.8000001907349px">acceptable but consider using 1+0 or RAID 5 the performance on the two </span><span style="font-size:12.8000001907349px">may vary depending on the RAID controller but 1+0 is generally faster.</span><br style="font-size:12.8000001907349px"><span style="font-size:12.8000001907349px">2.c) always optimize for the stripe which in theory should be </span><span style="font-size:12.8000001907349px">automatic if the RAID controller reports the required information to </span><span style="font-size:12.8000001907349px">do it, but in practice most do not so you should do it manually and </span><span style="font-size:12.8000001907349px">Red Hat's support site just added a nice tool to do it for you.</span></div><div>Comment :- What about if i will split my 24 SSD in to two virtual drive with RAID 6...actually i want to leverage the performance of the SSD..without wastage too much of the disk capacity and redundancy...<br style="font-size:12.8000001907349px"><br style="font-size:12.8000001907349px"><span style="font-size:12.8000001907349px">Thanks,</span></div><div><span style="font-size:12.8000001907349px">Punit</span></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Mar 24, 2015 at 11:01 AM, Paul Robert Marino <span dir="ltr"><<a href="mailto:prmarino1@gmail.com" target="_blank">prmarino1@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">A few things to clarify<br>
1) are all 4 for Gluster or are they the whole cluster.<br>
keep in mind that Ovirt expects that you are running Gluster on<br>
different nodes than your VM's. while I do not completely agree with<br>
that idea that is how it is now.<br>
<br>
2) RAID hardware is not necessarily a bad thing, in fact the cache may<br>
help significantly. that said it is not required.<br>
things to keep in mind if you are using RAID hardware for performance<br>
2.a) RAID 0 will give you the best speed if you are using 3 nodes for<br>
quorum. keep in mind this will require more work in the case of a<br>
drive failure to remake the brick.<br>
2.b) if you are only doing a mirror across 2 bricks RAID 0 may be<br>
acceptable but consider using 1+0 or RAID 5 the performance on the two<br>
may vary depending on the RAID controller but 1+0 is generally faster.<br>
2.c) always optimize for the stripe which in theory should be<br>
automatic if the RAID controller reports the required information to<br>
do it, but in practice most do not so you should do it manually and<br>
Red Hat's support site just added a nice tool to do it for you.<br>
<br>
3) for performance more RAM is rarely a bad thing; however it can be<br>
with sustained writes.If you are constantly writing faster than the<br>
disks can handle you will eventually fill the buffers and block till<br>
they flush. so the answer here is if you are mostly dealing with reads<br>
more and bursts of writes more ram is better. if you are dealing with<br>
constant writes it may cause uneven performance and may in fact be<br>
detrimental.<br>
<br>
4) On the 10Gb consider using consider using RDMA Iverbs over Ethernet<br>
it may help.<br>
<br>
5) Look closely at the scheduler on the disks the default these days<br>
is CFQ (Completely Fair Queuing). I find DEADLINE works better for<br>
Gluster; however depending on if you are using a RAID and you have a<br>
properly aligned file system you may find NOOP may provide better<br>
performance you need to do some testing with your hardware to<br>
determine if this is the case.<br>
<div><div class="h5"><br>
On Mon, Mar 23, 2015 at 9:43 PM, Punit Dambiwal <<a href="mailto:hypunit@gmail.com">hypunit@gmail.com</a>> wrote:<br>
> Hi,<br>
><br>
> I want to use Glusterfs with Ovirt 3.5...please help me to make the<br>
> architecture stable for the production use :-<br>
><br>
> I have 4 servers...every server can host 24 SSD disk(As bricks)..i want to<br>
> deploy distributed replicated storage with replica =2....i don't want to use<br>
> the Hardware RAID...as i think it will badly impact the performance...<br>
><br>
> 1. Glusterfs 3.5 or 3.6 ?? (which one will be stable for the production<br>
> use).<br>
> 2. Do i use the Hardware RAID or Not ??<br>
> 3. IF HW RAID then which RAID level and does it impact the performance...<br>
> 4. I want to make it rock solid...so it can use for production purpose...<br>
> 5. How much RAM should be sufficient on each server...on the each server i<br>
> have two E5 CPU's...<br>
> 6. For Network Connectivity i have 2*10G NIC with bonding on each server...<br>
><br>
> Thanks,<br>
> Punit<br>
><br>
</div></div>> _______________________________________________<br>
> Gluster-users mailing list<br>
> <a href="mailto:Gluster-users@gluster.org">Gluster-users@gluster.org</a><br>
> <a href="http://www.gluster.org/mailman/listinfo/gluster-users" target="_blank">http://www.gluster.org/mailman/listinfo/gluster-users</a><br>
</blockquote></div><br></div>