[cisco-voip] UCS C210 Replace 146 GB Disk in RAID5 with 300 GB Disk

[Lelio Fulgenzi]

Thanks for laying it out Ryan.


From: Ryan Ratliff (rratliff) [mailto:rratliff at cisco.com]
Sent: Wednesday, November 15, 2017 10:57 AM
To: Lelio Fulgenzi <lelio at uoguelph.ca>
Cc: Charles Goldsmith <wokka at justfamily.org>; cisco-voip list <cisco-voip at puck.nether.net>
Subject: Re: [cisco-voip] UCS C210 Replace 146 GB Disk in RAID5 with 300 GB Disk

The 4 arrays on BE7K also have an impact on IOPS for the volume, though the servers are primarily built that way for array rebuild times as has been noted previously in this thread.

On that server especially while one volume may be fine from a space perspective note that the amount of IOPS for DAS is derived from the number of disk spindles in the array.
This means it is entirely possible to cause IO starvation problems on a BE7K by putting all the VMs on a single storage volume.
Spreading your VMs across the arrays will provide you protection against multiple disk failures (because they are RAID 5) AND spread the IOPS load across them.

With respect to rebuilding a TRC’s RAID5 into RAID6 or RAID10 that server is no longer classified as a TRC and thus Cisco cannot guarantee the same level of performance we would otherwise.
You may be perfectly comfortable with making such a change and willing to take on the risk, but guaranteeing the server meets the IOPS requirements of the apps running on it is your responsibility, not ours.

If things go sideways you can expect TAC to ask you to rebuild it to a RAID5 (or at least one volume) the way it shipped as part of the troubleshooting process.

-Ryan

On Nov 14, 2017, at 11:33 AM, Lelio Fulgenzi <lelio at uoguelph.ca<mailto:lelio at uoguelph.ca>> wrote:

Yeah, I hear ya. I mean, it’s not like there are not advantages, but, still, the managing of which array to put things in. ugh.


From: Charles Goldsmith [mailto:wokka at justfamily.org]
Sent: Tuesday, November 14, 2017 10:48 AM
To: Lelio Fulgenzi <lelio at uoguelph.ca<mailto:lelio at uoguelph.ca>>
Cc: Ryan Huff <ryanhuff at outlook.com<mailto:ryanhuff at outlook.com>>; cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
Subject: Re: [cisco-voip] UCS C210 Replace 146 GB Disk in RAID5 with 300 GB Disk

Lelio, if you are using those be7k's for UC apps, your limiting factor is always the CPU cores, you won't run out of drive space.  Also, the be7K-h is a workhorse of a server, fastest one I've ever built a cluster on.  4 independent arrays means you can easily do maintenance on one app without affecting the performance of another, assuming you separate your apps between arrays and stagger them out.  I did 4 simultaneous installs on one, each app on it's own partition and they were all done in under 2 hours.

Sadly, the pricepoint on the H isn't there for most customers.  the M is nice, but only has 2 of the arrays, but I'll take an H for an install any day!

On Tue, Nov 14, 2017 at 9:34 AM, Lelio Fulgenzi <lelio at uoguelph.ca<mailto:lelio at uoguelph.ca>> wrote:

The idea of RAID drives and managing the different volumes always had me on the fence on how to do things. In a perfect world, I’d stick with one big RAID 6 array with a spare on the shelf.

The BE7K servers I ordered were delivered with 4 RAID 5 arrays. Personally, while I can appreciate separating the arrays, I don’t like losing that extra space and managing which volume to put images on is a pain.

I’ll admit, I looked at RAID 10 (when I was first reading the TRC specs) and was confused to heck. I did finally understand things after referring to a colleague, but it was a lot of drawing out.

I will say this, RAID isn’t gonna protect you if you don’t have platform monitoring on. You need to know the second a drive fails so you can proceed accordingly.

Also, if the ever do construction in your computer room, do yourself a favour, go to the hardware store, buy a 9.99 loose fibre furnace filter and stick it in front of your air intakes.



From: cisco-voip [mailto:cisco-voip-bounces at puck.nether.net<mailto:cisco-voip-bounces at puck.nether.net>] On Behalf Of Charles Goldsmith
Sent: Tuesday, November 14, 2017 9:48 AM
To: Ryan Huff <ryanhuff at outlook.com<mailto:ryanhuff at outlook.com>>
Cc: cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
Subject: Re: [cisco-voip] UCS C210 Replace 146 GB Disk in RAID5 with 300 GB Disk

I've seen one URE fail in a raid 5 resilvering process, years ago on a DG system.  Had to rebuild and restore from backup, fun times.

I agree Ryan, on a TRC system and RMA a drive, you stick with it.

From my reading on TRC, you can rebuild as a RAID 10 and get faster speeds, but you lose some space in the process.

On my personal systems, I'm using RAID 10 everywhere.

On Tue, Nov 14, 2017 at 8:17 AM, Ryan Huff <ryanhuff at outlook.com<mailto:ryanhuff at outlook.com>> wrote:
As I’ve read and understood; it isn’t due to actual functionality though. It is as you say, due mostly to longer rebuild times (indexing a physically larger geometry than the rest of the array members, for a smaller logical geometry) and the risk (rare IMO) to the rest of the array (as a rebuild will stress the array and could cause other, near-death disks to fail thereby causing the array to fail). It also wastes the extra horsepower of the disk since the existing RAID can’t capitalize on the resources of the larger disk.

So in a case of, would you go out and buy a new disk that way .... I’d say no; but if that is the result of a covered RMA, I’d say go for it.

I’m no diskologist though ... just based on my own experiences of what has worked for me for the last couple of decades ... and I’ve never lost a server ... outside of that one time when my pants pocket snagged the release on the 2nd disk in a R5 on my way out the door ... bad memories.
-Ryan

On Nov 14, 2017, at 9:03 AM, Charles Goldsmith <wokka at justfamily.org<mailto:wokka at justfamily.org>> wrote:
Keep in mind, RAID 5 is ok for smaller disks, but larger disks it's no longer recommended, but sadly, the best article about it is from Dell: http://en.community.dell.com/techcenter/b/techcenter/archive/2012/08/14/new-equallogic-raid-tech-report-considerations-and-best-practices-released

With bigger disks, it's even said that RAID 6 is no longer good enough, due to large rebuild times in case of a failure.  http://www.zdnet.com/blog/storage/why-raid-6-stops-working-in-2019/805

On Tue, Nov 14, 2017 at 7:41 AM, Ryan Huff <ryanhuff at outlook.com<mailto:ryanhuff at outlook.com>> wrote:
Reto,
Seek/rpm speeds and media type (flash, sata ... etc) are usually what matter the most for RAID disks. If your only difference is total storage capacity, the bigger disk will usually work just fine, your just gonna waste the additional 154GB of space (because the RAID will only provision 146GB of that 300GB disk).

Just remember on a RAID 5, don’t pull/lose more that 1 disk at a time .... painful lesson long ago I share over beer every now and then.

-Ryan

On Nov 14, 2017, at 8:23 AM, Reto Gassmann <voip at mrga.ch<mailto:voip at mrga.ch>> wrote:
Hallo

We have a UCS C210 Server with 10x146 GB Disks. One of the Disks failed and I got a 300 GB replacement Disk from Cisco.

Is that a problem if I replace the defect 146 Disk in the RAID 5 with a 300 GB Disk?

Thanks for help
Regards Reto
_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
https://puck.nether.net/mailman/listinfo/cisco-voip

_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
https://puck.nether.net/mailman/listinfo/cisco-voip



_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
https://puck.nether.net/mailman/listinfo/cisco-voip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/cisco-voip/attachments/20171115/9caddbda/attachment.html>