|
PCjr sidecar posted:You're doing this with a LSI RAID controller? You're going to hit its limit before you hit the capability of the SSDs. 400,000 IOPS is the limit of the raid controller in question. cheese-cube posted:Yeah LSI MegaRAID does things in a weird way compared to other more straight-forward controllers (i.e. Adaptec). It's a server from iXSystems. We're building a new direct attached storage virtualization cluster that will blow your balls off. edit: with redundant 10gbit we can live migrate 100GB VMs between servers if extremely quickly, if necessary. I mean, at 1-2GB/s ... you do the math. feld fucked around with this message at 16:06 on Jul 29, 2014 |
#
?
Jul 29, 2014 16:02
|
|
|
|
| # ? May 9, 2024 21:06 |
|
|
feld posted:Select all 24 drives and choose RAID1. You'd think it's just a giant RAID1, right? 500GB drives, you end up with only 500GB storage with a ton of redundancy. Turns out it's exactly the same volume size if it was a RAID 10 over 12 mirrors (500*12)! All volumes blink when doing writes, so it's striped across all of them. I would try this. Create 12 RAID 1 volumes with 2 drives each in raid 1, after making the 12 RAID 1 volumes, go back and create a raid 0 setup to stripe across the 12 RAID 1 volumes. feld posted:It's a server from iXSystems. We're building a new direct attached storage virtualization cluster that will blow your balls off. Wait, doesn't Dilbert As FUCK fucked around with this message at 16:09 on Jul 29, 2014 |
|
#
?
Jul 29, 2014 16:07
|
|
|
feld posted:400,000 IOPS is the limit of the raid controller in question. That's with a single controller and 8 SSDs in RAID0. I hope you bought more controllers. Edit: Dilbert take your meds. Pile Of Garbage fucked around with this message at 16:12 on Jul 29, 2014 |
|
#
?
Jul 29, 2014 16:09
|
|
|
cheese-cube posted:That's with a single controller and 8 SSDs in RAID0. I hope you bought more controllers. We don't need to go beyond the 400,000 IOPS limitation, so that's not a concern. Dilbert As gently caress posted:I would try this. Citrix Xenserver. I wish it was FreeBSD based, but it's not. Again, we can't do 12 RAID1 volumes because the controller caps out at 8 drives in a single volume and 8 total volumes. 
|
|
#
?
Jul 29, 2014 16:12
|
|
|
feld posted:We don't need to go beyond the 400,000 IOPS limitation, so that's not a concern. Oh ha, misread the 8 total volumes. Derp, yeah 3 raid 10 with 8drives would not be ideal, but it would give you the same total capacity as raid 10 but not on an individual datastore level. Are 2TB volumes too small for what you are trying to do?
|
|
#
?
Jul 29, 2014 16:19
|
|
|
Dilbert As gently caress posted:Oh ha, misread the 8 total volumes. Derp, yeah 3x8 raid 10 drives would not be ideal, but it would give you the same total capacity as raid 10 but not on an individual datastore level. Are 2TB volumes too small for what you are trying to do? This has been an internal debate of mine. I'd love to not have multiple local storage volumes in case we have anyone beg for large amounts of disk space, but I dont think I can have my cake and eat it too
|
|
#
?
Jul 29, 2014 16:21
|
|
|
feld posted:This has been an internal debate of mine. I'd love to not have multiple local storage volumes in case we have anyone beg for large amounts of disk space, but I dont think I can have my cake and eat it too Couldn't you solve this with with adding another virtual hard disk to the VM on a separate datastore, then doing an internal raid or extent within the guest? Otherwise, you could do, say 2 RAID 10's with 8 drives, then a RAID 5 with 8x drives giving you 3.5TB; while leaving the the 25th drive as a spare for anything that needs it. This would give you some flexibility on the datastore size if it ever became a problem, and while RAID 5 isn't IDEAL for SSD's it does work, and you probably won't notice the write penalty. Dilbert As FUCK fucked around with this message at 16:29 on Jul 29, 2014 |
|
#
?
Jul 29, 2014 16:25
|
|
|
Anyone know how to get IBM on the phone without waiting for a callback for SAN issue? loving waiting for a callback.
|
|
#
?
Jul 29, 2014 17:46
|
|
|
feld posted:400,000 IOPS is the limit of the raid controller in question. And you've got ~1-2M IOPS worth of SSD behind it. Which is fine if: feld posted:We don't need to go beyond the 400,000 IOPS limitation, so that's not a concern. but my original statement still stands. goobernoodles posted:Anyone know how to get IBM on the phone without waiting for a callback for SAN issue? loving waiting for a callback. Call your sales rep/account manager. (This only works if you spend enough with IBM.)
|
|
#
?
Jul 29, 2014 17:50
|
|
|
feld posted:We don't need to go beyond the 400,000 IOPS limitation, so that's not a concern. This seems wildly optimistic for a real world performance estimate. Hope all of your data is small block read or write only synthetic IO.
|
|
#
?
Jul 29, 2014 17:53
|
|
|
So it looks like Veeam backups last night filled up our SAN and locked up our hosts. Trying to get IBM on the horn, but has anyone dealt with this situation before? How can I delete these failed snapshots to get the hosts responsive again? 
|
|
#
?
Jul 29, 2014 18:12
|
|
|
goobernoodles posted:So it looks like Veeam backups last night filled up our SAN and locked up our hosts. Trying to get IBM on the horn, but has anyone dealt with this situation before? How can I delete these failed snapshots to get the hosts responsive again? Yeah you probably want to be calling Veeam, not IBM.
|
|
#
?
Jul 29, 2014 18:17
|
|
|
goobernoodles posted:So it looks like Veeam backups last night filled up our SAN and locked up our hosts. Trying to get IBM on the horn, but has anyone dealt with this situation before? How can I delete these failed snapshots to get the hosts responsive again? For a temp fix, you can set the reservation for memory of virtual machines equal to the allocated memory. It is a temporary fix, but it will remove the .vswap files allowing you to get a few things up and running for the short term. But yeah like cheese said, call veeam. oh hey look new avatar
|
|
#
?
Jul 29, 2014 18:22
|
|
|
^^ lolgoobernoodles posted:Anyone know how to get IBM on the phone without waiting for a callback for SAN issue? loving waiting for a callback. http://www-304.ibm.com/support/customercare/sas/f/handbook/getsupport.html#9 Ask to speak to a duty manager. If that doesn't help, ask to open a critical situation.
|
|
#
?
Jul 29, 2014 19:41
|
|
|
goobernoodles posted:Anyone know how to get IBM on the phone without waiting for a callback for SAN issue? loving waiting for a callback. This way you have time to update to the latest firmware, like they will ask you when they call 
|
|
#
?
Jul 29, 2014 19:57
|
|
|
Nice avatar. It doesn't look like the SAN didn't actually fill up. I think one of the LUN's changed it's preferred path and the hosts weren't able to see it on the new path and became unresponsive. Ran IBM's "Redistributed logical drives" utility and everything came back up. God drat I forgot how much I loving hate IBM support. They look for a reason to get off the phone from the moment you start talking to them. Don't get me started on IBM and their firmware updates.
|
|
#
?
Jul 29, 2014 20:04
|
|
|
goobernoodles posted:Nice avatar. What model are you using if you don't mind me asking?
|
|
#
?
Jul 29, 2014 20:08
|
|
|
goobernoodles posted:Don't get me started on IBM and their firmware updates.
|
|
#
?
Jul 29, 2014 20:26
|
|
|
cheese-cube posted:What model are you using if you don't mind me asking? The_Groove posted:We run firmware on a ton of netapps that's approved/tested by IBM's cluster team, but is generally old enough that their website for firmware downloads has aged off the version we need by the time it's been approved. It happened a second time. This time I told them what happened on the other server and specifically asked if we needed to hop versions. The dude assured me we didn't have to. Bricked the primary UEFI*. Ugh. goobernoodles fucked around with this message at 04:02 on Jul 30, 2014 |
|
#
?
Jul 30, 2014 00:24
|
|
|
Only IBM.
|
|
#
?
Jul 30, 2014 03:36
|
|
|
goobernoodles posted:DS3300 with an EXP3000 expansion. Full with 300 and 600gb SAS drives. My Equallogic is on a 6 year old firmware. Dell told me not to bother updating if everything is working. There are a ton of bug fixes but we clearly are not running into them, and they haven't tested a jump that large. Doing incremental for things Dell has tested and would support would take so much scheduled maintenance (we have a single array running everything) it would just not be worth it.
|
|
#
?
Jul 30, 2014 03:54
|
|
|
I can tell you that there is a ton of cool new poo poo that was implemented in the past 6 years on EqualLogic, not to mention the bug fixes. I would take the time to get it upgraded because if you ever do an upgrade you can add the new array, join it to the group, and then migrate the old one out with 0 downtime. Never had any problems doing EqualLogic firmware updates. Just do a few at a time every couple weekends. Not to mention the fact that you probably should be planning an upgrade soon anyways. Running on 6 year old hardware is asking for trouble and I can't imagine the warranty renewal on that thing is cheap.
|
|
#
?
Jul 30, 2014 03:58
|
|
|
I typically don't upgrade the firmware on my arrays unless there is a compelling reason to do so. I think if we hadn't have hit some hardware failures we would be pushing 30 months of uptime on our main production netapp filers.
|
|
#
?
Jul 30, 2014 04:29
|
|
|
I don't do them either unless the vendor has one of those "install this upgrade or you will lose all your data at some random future date" firmware releases. We have some equallogics that were last rebooted in Feb 2012 that are still chugging along just fine.
|
|
#
?
Jul 30, 2014 16:08
|
|
|
IT wasn't maintained at all for the last 5-6 years here so that explains the storage. Background: I work for a city government, IT was handled horribly for the 4 years prior to my boss coming in (read the previous person did absolutely nothing and it too that long to get her fired, she had no staff and did 100% of labor by contract). My boss used contractors for the first year to maintain and figure out what was going on, we have zero documentation that we haven't created. My boss hired people straight out of school to basically run everything in IT in the city. She clearly has an eye for talent everyone seems to be excelling and figuring stuff out. Everything is a ton better but we need more, and faster storage at this point to continue with the fixes. As a note: We have 500GB of SAN space left, are on AD functionality 2003, with a single level domain, we currently want to setup a proper domain and get it on functionality level 2012 and migrate everything, but to do that we need storage which is now my responsibility. We don't have a storage person. I want to get into storage as a focus in my career though and have the ability to touch storage here. I've been recently tasked with getting new storage entirely. Going with equallogic again is an obvious option, but Nimble looks pretty promising. We've had bad luck with our HPs at another site causing the VMs to go offline when Veeam does backups and everyone is pointing the finger and we don't pay anyone enough for them to just fix it. I was trying to hold off on the new thread and hoped it would answer a few questions about vender choices (the current OP seems pretty dated). Also wondering if any certs are worth getting in storage, I only see vender specifics and Storage+ would studying for Storage+ be a good idea? Is it respected in the field, or is something else respected more? We are running Exchange and due to how police email has to be handled, it's unlikely that we can move to gmail, so IOPS a reasonable price matters. Also something an idiot can manager when I leave here, I don't plan on sticking around for the life of the device.
|
|
#
?
Jul 30, 2014 16:31
|
|
|
If you don't want a storage admin and want something very simple, high performing, and cost effective, nimble is a solid win. They have a bunch of case studies working with government data and how simple it is to setup and maintain. The problem that some companies see with nimble is that it is a smaller company, and vendor/MSP support isn't as wide spread as say, equal logic, 3Par, or EMC. Aside from Nimble, check out hybrid arrays, where they leverage caching to provide cost effective IOPS. As far as a Cert goes, I would vouch for the EMCSIA. While the book is a bit of a dry read, it does provide you with a bunch of useful knowledge. I think I have a free 45 day trial to the book I never used, might still be good if you want it. Have you had capacity planner on your environment, or have a guestimation of what kind of IO you are looking at?
|
|
#
?
Jul 30, 2014 16:47
|
|
|
Seconding Nimble for the dead simple storage to manage. It is loving boring.
|
|
#
?
Jul 30, 2014 17:13
|
|
|
NippleFloss posted:This seems wildly optimistic for a real world performance estimate. Hope all of your data is small block read or write only synthetic IO. It doesn't matter what the VMs do for block size read or write as the hypervisor only operates at a certain fixed block size. You tune for the hypervisor, not the VMs inside. feld fucked around with this message at 17:17 on Jul 30, 2014 |
|
#
?
Jul 30, 2014 17:14
|
|
|
feld posted:It doesn't matter what the VMs do for block size read or write as the hypervisor only operates at a certain fixed block size. You tune for the hypervisor, not the VMs inside. This is completely wrong, but even if it wasn't wrong that wouldn't mean that VMWare IO is in any way similar to a synthetic benchmark. pixaal posted:We are running Exchange and due to how police email has to be handled, it's unlikely that we can move to gmail, so IOPS a reasonable price matters. Newer versions of Exchange are very low IO and very latency tolerant. Starting with 2010 it's mean to run on SATA. You don't need fast storage for Exchange. In fact, you'll see storage vendors brag about how many mailboxes they can get into whatever array or cluster and the numbers sound pretty impressive until you realize that most of that is due to Exchange becoming very very disk efficient. The CAS servers handle the majority of the latency sensitive mail operations provided they are sized properly with enough memory. adorai posted:I typically don't upgrade the firmware on my arrays unless there is a compelling reason to do so. I think if we hadn't have hit some hardware failures we would be pushing 30 months of uptime on our main production netapp filers. After having been bitten badly by some bugs after recommending an upgrade to 8.1 I'm in this camp too. If it's working fine and you don't need the new features then don't upgrade just for the sake of doing it. If there's a really major bug discovered in that code line I'll recommend the P release that fixes it, but not a major upgrade without a compelling reason.
|
|
#
?
Jul 30, 2014 17:45
|
|
|
I'm not sure if this is the right place for this, but I'll ask anyway I suppose. So I work in IT at a private school, and we're migrating most of our stuff to cloud based storage (we chose enterprise dropbox for most of the admin accounts, as we like their encryption, price, and ease of use/implementation). The only problem we have now is that we have a large repository of pictures/videos from all our various sporting events, plays, awards ceremonies, blah blah, and nowhere to store them in the cloud thats also easy to access. The problem with dropbox is that if you install the client on the computer, and the photo share is in the folder, it downloads the entire library to the users PC in order to sync. For users with smaller laptops and such, it would completely fill the hard drive. We originally had this just on the server in a random place, which worked in the building, but we have people that need access from home, an a VPN is "too complicated" for some of our users. What would the best storage solution be for a large amount of multimedia, easily shared between multiple users, both in the building itself, and also outside of our network?
|
|
#
?
Jul 30, 2014 21:28
|
|
|
Dilbert As gently caress posted:If you don't want a storage admin and want something very simple, high performing, and cost effective, nimble is a solid win. They have a bunch of case studies working with government data and how simple it is to setup and maintain. The problem that some companies see with nimble is that it is a smaller company, and vendor/MSP support isn't as wide spread as say, equal logic, 3Par, or EMC. Aside from Nimble, check out hybrid arrays, where they leverage caching to provide cost effective IOPS. The Equallogic UI displays nothing that I can find about IOPS so not really too much other then a a decent number of exchange mailboxes, a few dozen SQL databases on VMs, but no where near as heavy use as most people would assume when you say SQL. Its mostly 1 or 2 people in a department using a single piece of software once a month. Our large city wide DBs are access, which is fun. We are currently at just under 3TB, and don't have any immediate plans to expand beyond setting up a second domain and starting to migrate everything (FQDN is wrong, no support for Server 2008 R2 because of it). I'm going for 12TB of storage from a single array, and preferably something that will upgrade easily. Is there an easy way for me to test what my current EqualLogic array is capable of in IOs, or what is attempting to be written? We are using VMware vCenter if its going to be easier to tell demand from the Hypervisor. I'd prefer to buy something for what we need instead of guess on IOPS. I'm only a good tenth of the way into a book on storage so if there is an obvious way of calculating it I haven't reached that chapter sorry for the silly question. Really I didn't plan on asking until I had finished reading a book but my boss kind of wants more storage. She thought we had another year of support and we don't so the array is just a timebomb at this point. Renew support or replace, and we are using most of the space on it, and are pretty sure we are using every IO the thing can put out for most of the day.
|
|
#
?
Jul 30, 2014 23:30
|
|
|
pixaal posted:Is there an easy way for me to test what my current EqualLogic array is capable of in IOs, or what is attempting to be written? What do you have again? Dell was more than happy to provide us with a spreadsheet of max IOPS for each of their current units with various disk configs.
|
|
#
?
Jul 30, 2014 23:54
|
|
|
pixaal posted:The Equallogic UI displays nothing that I can find about IOPS so not really too much other then a a decent number of exchange mailboxes, a few dozen SQL databases on VMs, but no where near as heavy use as most people would assume when you say SQL. Its mostly 1 or 2 people in a department using a single piece of software once a month. Our large city wide DBs are access, which is fun. You may want to bring in a VAR/MSP to do a storage analysis, or look up running capacity planner for your environment. Which if you are looking for backup solutions Avamar is great, if you don't want VDP-A for some reason. Like nipple said Exchange is built to run on SATA, the IOPS are not as important as the allocated space for the servers, since people love 50G mailboxes. quote:We are currently at just under 3TB, and don't have any immediate plans to expand beyond setting up a second domain and starting to migrate everything (FQDN is wrong, no support for Server 2008 R2 because of it). I'm going for 12TB of storage from a single array, and preferably something that will upgrade easily. What makes up that 3TB is it 24 146GB 15K drivesor just a bunch of sata thrown at something? A C210 from nimble would probably suite your needs perfectly. Since I have no idea what your budget is I'd say look at Nimble, EMC, and Dell. While dell isn't the best they do offer offer quite a bit of nice things at a low cost, the MD3620i with SSD cache is amazing. quote:Is there an easy way for me to test what my current EqualLogic array is capable of in IOs, or what is attempting to be written? We are using VMware vCenter if its going to be easier to tell demand from the Hypervisor. I'd prefer to buy something for what we need instead of guess on IOPS. I'm only a good tenth of the way into a book on storage so if there is an obvious way of calculating it I haven't reached that chapter sorry for the silly question. Really I didn't plan on asking until I had finished reading a book but my boss kind of wants more storage. She thought we had another year of support and we don't so the array is just a timebomb at this point. Renew support or replace, and we are using most of the space on it, and are pretty sure we are using every IO the thing can put out for most of the day. http://blog.synology.com/blog/?p=2225 This dude says it better than I can. You can do it through VMware probably easier than Dell's equal logic. Aside from that capacity planner or call up dell, ask about DPACK and they will give you a report so long as you ask for a quote. http://www.nimblestorage.com/customers/ Check out some of nimble's case studies. As far as the book which one are you reading, I am curious for my own knowledge. Also might I ask; What is your backup solution and how are you looking to replace or assimilate it to your new storage? Dilbert As FUCK fucked around with this message at 23:59 on Jul 30, 2014 |
|
#
?
Jul 30, 2014 23:56
|
|
|
pixaal posted:I've been recently tasked with getting new storage entirely. 1) NetApp: Probably the most expensive of the three, but certainly also the most feature rich. The tools that integrate with your application backups are amazing, and really sell the box for me. Consider the savings from being able to leverage these backup tools in any purchase decision. Also remember this box will do any protocol you want. NFS, iSCSI, FC, and CIFS are all built in. 2) Nimble: You'll hear from the sales guys that, "they have a great story." It might be true. The reality is that if you want a screaming fast box for cheap, and you can live with iSCSI only, this is a great buy. 3) Oracle: Oracle bought Sun, the company that created ZFS and pioneer hybrid flash and disk pools. The hardware is inexpensive, maintenance is very reasonable, and the boxes work quite well. Like Nimble they leverage inline compression of data to both save space and cut down on IOPS and they put a shitload of cache in front of both reads and writes. Like NetApp it supports pretty much every protocol you could want. They don't however have all the cool tools. Also keep in mind that you might want two of whatever you buy. SAN to SAN replication is simply a great feature and one that we couldn't live without. As a municipal user, you might want to consider finding another government entity to partner with for replication and DR. Hell, you may even have a local business that uses NetApp that would let you purchase some disks and plug into their array for backups. I know that if my local city government approached me with such a request I would look for a way to accommodate them. Banks especially might be interested, as they could potentially use it for the Community Reinvestment Act program.
|
|
#
?
Jul 31, 2014 00:07
|
|
|
Pop quiz hotshots: If you were buying a new NetApp and were going to have 44 data disks available for an aggregate, would you go with option 1, 2, or 3? option 1) 2 raid groups of 21 disks and two hot spares option 2) 3 raid groups of 14 disks and two hot spares option 3) some variation of one of the above, but with an aggregate just for the root volume
|
|
#
?
Jul 31, 2014 00:10
|
|
|
pixaal posted:The Equallogic UI displays nothing that I can find about IOPS If you are using Equallogic, you should REALLY be running SAN headquarters. It gives you all of this sort of information. You can download it from your support portal and it's pretty much a requirement for later firmwares (you'll get a persistent alert in the EQL management console if you don't have the group hooked up to SAN headquarters.)
|
|
#
?
Jul 31, 2014 00:11
|
|
|
adorai posted:Pop quiz hotshots: If you were buying a new NetApp and were going to have 44 data disks available for an aggregate, would you go with option 1, 2, or 3? 7 mode or clustered? What size disks?
|
|
#
?
Jul 31, 2014 00:36
|
|
|
NippleFloss posted:7 mode or clustered? What size disks?
|
|
#
?
Jul 31, 2014 00:40
|
|
|
adorai posted:Important information I suppose. Clustered mode, 900GB SAS 10K or 15K Also what kind of IOPS am I looking for and what level of redundancy? Prod DB's, VDI, etc adorai posted:Obviously I do not know all the details of your shop, but based on what you have posted I would be looking at three vendors. also why the bashing of post processed ISCSI? It owns. Dilbert As FUCK fucked around with this message at 00:49 on Jul 31, 2014 |
|
#
?
Jul 31, 2014 00:46
|
|
|
|
| # ? May 9, 2024 21:06 |
|
|
Dilbert As gently caress posted:10K or 15K C DSK SHLF,24x900GB,6G,0P,-C I don't know the exact number of IOPS but it is a mixed workload including SQL, exchange, VMware, and CIFS. No VDI. Dilbert As gently caress posted:also why the bashing of post processed ISCSI? It owns. adorai fucked around with this message at 00:51 on Jul 31, 2014 |
|
#
?
Jul 31, 2014 00:49
|
|