|
NippleFloss posted:It's pretty hard to gently caress up a FAS install short of mis-cabling or really suboptimal aggregate layouts. I'd be curious to know what exactly is wrong with it if it's been in production for years. Feel free to PM if you'd rather respond privately. I do 7-mode/cDOT PS for a NetApp Partner.. I've seen some "special" installs over the years that I've fixed, and I'd be curious to know too  A few customer "I know what I'm doing, I'll self install" that worked, until it didn't. Like plugging circle to circle, square to square on SAS connectors.. or using DS14s, and creating an aggregate of all the disk 0s (except the first enclosure), another of all the disk 1s (except the second enclosure), etc.. so 14 aggregates. Now, when they had a shelf failure, it wasn't a problem for them.. but that's certainly not a good way to get the best disk utilization possible. My favourite though is a FAS2240 that apparently panic'ed while running 8.1.1RC1, causing customer exchange environment to go offline. They called support, and it apparently took support 4 hours to get the system back online. We'd just taken this customer on, so we offer to do a post-incident evaluation. Go in and look at the /etc/rc file. Network ifgrp definitions are different on each head, partner statements where present aren't valid, and cf wasn't enabled. Generally poo poo work. I was surprised to see these still there, since after 4 hours on the phone to NetApp, you'd think they would have fixed that..
|
#
?
Aug 16, 2014 20:09
|
|
|
|
| # ? May 26, 2024 22:43 |
|
|
theducks posted:I do 7-mode/cDOT PS for a NetApp Partner.. I've seen some "special" installs over the years that I've fixed, and I'd be curious to know too It did end up being a cabling issue, though a somewhat obscure one, which is why it took it more than a year to bite them. The install was done by NetApp or a NetApp partner, so evidently the installer just didn't fully read his documentation because the mistake he made is called out in there.
|
|
#
?
Aug 16, 2014 20:51
|
|
|
I can just run the installer and update Dell's Equalogic SAN HQ right? No need to shut anything down or whatever?
|
|
#
?
Aug 18, 2014 19:43
|
|
|
itskage posted:I can just run the installer and update Dell's Equalogic SAN HQ right? No need to shut anything down or whatever? Correct. Just watch for alerting or monitoring, if you have anything setup for that.
|
|
#
?
Aug 18, 2014 19:46
|
|
|
Got it. Thanks for the reply.
|
|
#
?
Aug 18, 2014 19:48
|
|
|
This made for a fun afternoon, still not sure what happened really, none of the VM's seem to have been doing anything unusual at the time. The latency was so bad their phone vm started dropping calls. It recovered by itself after about 30 minutes, nothing actually went down. Hoping equallogic can figure it out. sanchez fucked around with this message at 14:58 on Aug 25, 2014 |
|
#
?
Aug 18, 2014 22:50
|
|
|
sanchez posted:This made for a fun afternoon, still not sure what happened really, none of the VM's seem to have been doing anything unusual at the time. The latency was so bad their phone vm started dropping calls. It recovered by itself after about 30 minutes, nothing actually went down. Hoping equallogic can figure it out.
|
|
#
?
Aug 18, 2014 23:52
|
|
|
Had a fun little support story with Nimble today. Was doing an update on an array for 1.4.x to 2.x and the said array didn't have autosupport communication for a good chunk of time. This array was blocked from said update until support did a "fix". They had some bug where the firmware needed to be at a minimum level before it was updated to the latest. Their fix was to ssh into the array and run a touch command on a random rear end file. Without either this file or a recent timestamp, the pre-update check would fail. Not the most fancy way to do it, but seemed to work. They really don't trust their customers to handle poo poo on their own, and rather their support to handle it. Not a bad thing for a lot of people, but it does kinda suck when I can read and do a tiered update. Still loving these things.
|
|
#
?
Aug 20, 2014 04:02
|
|
|
Moey posted:Had a fun little support story with Nimble today. Was doing an update on an array for 1.4.x to 2.x and the said array didn't have autosupport communication for a good chunk of time. This array was blocked from said update until support did a "fix". They had some bug where the firmware needed to be at a minimum level before it was updated to the latest. Their fix was to ssh into the array and run a touch command on a random rear end file. Without either this file or a recent timestamp, the pre-update check would fail. Can't wait to start the implementation of what we were approved for. Furiously  ing our PR system in the hope that it goes through soon.
|
|
#
?
Aug 20, 2014 06:05
|
|
|
My organisation (small to midsize law firm) is implementing a new Practice Management application. We are purchasing new virtualization hosts and a dedicated SAN for this application. I could do with some advice on the SAN. Space-wise we need about 20TB usable. The software vendor uses the Microsoft SQLIO utility to size storage performance, and they have advised that we need a SAN that can meet the following benchmark, as observed by SQLIO: 64k test: 2625 random writes / 5500 random reads 8k test: 4465 random writes / 8250 random reads I don't know if the use of SQLIO is unusual or not, but that is the data I have. We've been to our normal suppliers for recommendations. One has suggested a NetApp FAS802A. The other has suggested a 3PAR Storeserv 7200. I am waiting on the full details and pricing. From the sounds of things, the HP will give us a lot more spindles and capacity for the price. We don't use SAN based replication or backups (we use Zerto for replication and Dell Appassure for backups), so we don't have a particular requirement for any of these features, although we would like cross-shelf HA (might be wrong terminology). Currently all our infrastructure is 1GBE. I know that 10GBE is becoming (?) popular but I don't know whether this is worth the cost premium. Any input is appreciated.
|
|
#
?
Aug 20, 2014 16:36
|
|
|
gallop w/a boner posted:My organisation (small to midsize law firm) is implementing a new Practice Management application. We are purchasing new virtualization hosts and a dedicated SAN for this application. I could do with some advice on the SAN. SQLio is a pretty good benchmark. You are one step ahead with a software supplier that delivers a baseline compared to most software companies. Consider yourself lucky  As for a SAN, your requirements are not that weird. I would take a look at the Netapp, their SQL backup software is amazing. If you are not going to use the Netapp tooling than pretty much any SAN will fit your needs. 3Par is nice what I have seen of them. I would ask for a couple of demonstrations and see for yourself which kit you like best. If you could swing it: For 20TB I would try an all flash array like a Pure and not have to worry about iops ever again. Mr Shiny Pants fucked around with this message at 17:33 on Aug 20, 2014 |
|
#
?
Aug 20, 2014 17:29
|
|
|
gallop w/a boner posted:Space-wise we need about 20TB usable. The software vendor uses the Microsoft SQLIO utility to size storage performance, and they have advised that we need a SAN that can meet the following benchmark, as observed by SQLIO: SQLIO is a pretty standard tool to use for benchmarking SQL performance, however it is rapidly becoming obsolete because the results it provides for a lot of modern storage arrays are basically meaningless. It writes empty blocks (all zeros) when it performs it's write testing and many modern arrays have things like inline compression, which is wildly effective on an all-zero, meaning that you drive almost no actual load on the back end disks during writes, and the entire workload will fit comfortably into cache during reads because it's only a few hundred MBs. Even if they don't do compression specifically a some arrays, like Compellant, will simply not write the zero data at all, basically doing very limited zero block only inline deduplication. The upshot is that SQLIO measurements can give incredibly unrealistic estimates of throughput depending on very specific implementation details around how the array handles empty writes. If you just need a big bucket of IOPs and don't really care about features then just about any vendor should meet your requirements and you should probably just go with the cheapest. But you should use this as an opportunity to consider what you like and dislike about your current backup and DR methodology and determine whether there are benefits to doing things differently, then let those discussions guide which features you would like to see in your storage.
|
|
#
?
Aug 20, 2014 17:56
|
|
|
Mr Shiny Pants posted:
and never worry about $110k out of your pocket.
|
|
#
?
Aug 20, 2014 20:03
|
|
|
Nitr0 posted:and never worry about $110k out of your pocket. I had no clue they were that expensive. IF he could swing it! Would be well spent though
|
|
#
?
Aug 20, 2014 21:41
|
|
|
Pay me 110k I'll find you a solution that works and then pocket the rest. Thx
|
|
#
?
Aug 20, 2014 22:37
|
|
|
Thanks for the input guys. It's always nice to have some advice from a neutral. Budget for the SAN is about 70K GBP so I don't think we can swing the all-flash array unfortunately.
|
|
#
?
Aug 21, 2014 11:26
|
|
|
Nimble, 3PAR, HDS, Compellent - tons of options and there isn't going to be a single "right" answer.
|
|
#
?
Aug 21, 2014 18:30
|
|
|
gallop w/a boner posted:Thanks for the input guys. It's always nice to have some advice from a neutral. Budget for the SAN is about 70K GBP so I don't think we can swing the all-flash array unfortunately. Niiiimmmmmbbbblllleeeee
|
|
#
?
Aug 21, 2014 18:52
|
|
|
gallop w/a boner posted:Thanks for the input guys. It's always nice to have some advice from a neutral. Budget for the SAN is about 70K GBP so I don't think we can swing the all-flash array unfortunately. Actually, you probably could get into a AFA a that price. Since you don't seem to care about features you might check out the EF series from NetApp. Pretty basic, low-frills, fast flash arrays. Not that you actually need flash to support that workload, and basically any vendor will be able to do it and probably within your price range. Your IO requirements didn't list any latency though, which makes it tough. You could do those IOPs at 50ms or at 1ms, and there's going to be a huge difference between the two. You're going to want to get clarification from the application developer on latency requirements.
|
|
#
?
Aug 21, 2014 20:21
|
|
|
Quick question for you NetApp experts. Last night, one of the disks on my second shelf (all 1T drives) failed. The system did not have a spare in that same size, so it used one of the 2T spares that was on another shelf to do data reconstruction. I replaced the failed 1T drive today. Is there a way I can easily release the 2T drive back as a spare and have the filer use the newly replaced 1T drive to do data reconstruction?
|
|
#
?
Aug 22, 2014 17:55
|
|
|
trilljester posted:Quick question for you NetApp experts. This is what the "disk replace" command is for.
|
|
#
?
Aug 22, 2014 18:53
|
|
|
adorai posted:The only times I have seen poo poo like that it's when a disk goes into heroic recovery and the OS waits for it. Equallogics response was unsatisfactory and we've lost a lot of confidence in them, Nimble CS300 ordered. 
|
|
#
?
Aug 25, 2014 14:57
|
|
|
Welcome to the darkside.
|
|
#
?
Aug 25, 2014 17:30
|
|
|
Not sure if I should post this in the Windows server thread or here but... Anyone have a software recommendation or method to bidirectionally sync a couple of shares from a windows file server to a another SMB share? I'm looking into potentially implementing Egnyte which requires your files to be on their pre-built VM to sync to their cloud location. Since not everyone in my company is going to need this, I was thinking of getting a relatively cheap ReadyNAS and sync our file server shares to it, which would in-turn sync to the cloud location. Just need something to keep the existing file server and the NAS in sync locally.
|
|
#
?
Aug 25, 2014 18:10
|
|
|
goobernoodles posted:Not sure if I should post this in the Windows server thread or here but... Anyone have a software recommendation or method to bidirectionally sync a couple of shares from a windows file server to a another SMB share? I'm looking into potentially implementing Egnyte which requires your files to be on their pre-built VM to sync to their cloud location. Since not everyone in my company is going to need this, I was thinking of getting a relatively cheap ReadyNAS and sync our file server shares to it, which would in-turn sync to the cloud location. Just need something to keep the existing file server and the NAS in sync locally. Any reason why you can't use DFSR?
|
|
#
?
Aug 25, 2014 18:25
|
|
|
cheese-cube posted:Any reason why you can't use DFSR? e: Definitely should have posted this in the Windows thread. Whoops.
|
|
#
?
Aug 25, 2014 18:51
|
|
|
What's the go-to open source file server distro? Is it still a toss up between Openfiler & FreeNAS? I'm looking specifically to run a fair number of VMware VMs in a test environment off of a 24 disk Supermicro disk array, all connected through iSCSI. I've been running a similar setup for about a year and a half with no hiccups. Performance isn't really a requirement, just stability/ease of use & simplicity of setup.
|
|
#
?
Aug 26, 2014 23:55
|
|
|
Wicaeed posted:What's the go-to open source file server distro? Is it still a toss up between Openfiler & FreeNAS?
|
|
#
?
Aug 27, 2014 00:04
|
|
|
Here's a funny one. My old EMC AX-100i (that's been out of production for a long time but still burns coal in the rack b/c I'm lazy) sent this message:quote:Event 2086 has occurred on storage system N/A for device N/A at 09/02/14 17:42:32 You'd never know anything was wrong by the lights on the box. Logging in shows that it's in crybaby mode about it's pet UPS which has been dead for years. Not that I need any more reasons to hate this machine, but does it really run Windows? 
|
|
#
?
Sep 4, 2014 16:49
|
|
|
NullPtr4Lunch posted:Here's a funny one. My old EMC AX-100i (that's been out of production for a long time but still burns coal in the rack b/c I'm lazy) sent this message: I'm not sure if it's any different on the VSP, but Hitachi's 9990, which was their to of the line array, team windows for the service processor. Even better, the service processor was really just a Windows laptop tucked away in a slot hidden behind a panel. That said, it had nothing to do with data processing, it was just for hardware monitoring, alerting, and running the management interface. I'd imagine your EMC box is doing the same thing, and the Windows portion isn't actually a part of the data plane.
|
|
#
?
Sep 4, 2014 17:35
|
|
|
Hey guys, looking at the HP MSA 2040 datasheet - saw this:quote:Choice of 8Gb/16Gb FC, 1Gb/10GbE iSCSI and 12Gb SAS to match the configuration needs of infrastructure. What's the difference between the 12Gb SAS controller and the others? What would the SAS controller be used for in a storage environment? Expansion?
|
|
#
?
Sep 11, 2014 16:16
|
|
|
Not a SAN, but Dell released a thing today that has some impact on storage density. R630 server. They added a 1.8" backplane option now. So, their 1U servers can have 24 1.8" hot swap SSD drives now. That's some crazy density. I put together some spitball configs. You could buy a server with 24 cores, 256gb of ram, and 4.8tb (raw) of flash storage for right about $20k. All in 1U.
|
|
#
?
Sep 11, 2014 16:29
|
|
|
bull3964 posted:Not a SAN, but Dell released a thing today that has some impact on storage density. Too bad it's from Dell. Recently switched to them from HP and it's been nothing but headaches. As for content, picked up 2 IBM v7000's with dual compression cards and 4TB of SSD, on top of 256GB of onboard flash cache. These things are screaming. They are doing the front end for 200TB of 10K disk. Been extremely pleasantly surprised using them how nice they are.
|
|
#
?
Sep 11, 2014 16:36
|
|
|
bull3964 posted:Not a SAN, but Dell released a thing today that has some impact on storage density. It probably ties back to their Nutanix relationship and this will used as an R630xd nutanix appliance soon.
|
|
#
?
Sep 11, 2014 16:40
|
|
|
Nukelear v.2 posted:It probably ties back to their Nutanix relationship and this will used as an R630xd nutanix appliance soon. The M630 also does 4x1.8" SSD's. That's pretty cool.
|
|
#
?
Sep 11, 2014 17:23
|
|
|
sudo rm -rf posted:Hey guys, looking at the HP MSA 2040 datasheet - saw this: Usually that's for direct attached storage, as opposed to accessing it over the network as you would with the other options.
|
|
#
?
Sep 11, 2014 17:30
|
|
|
sudo rm -rf posted:What's the difference between the 12Gb SAS controller and the others? What would the SAS controller be used for in a storage environment? Expansion? SAS is used for SAN back-end disk shelf loops and related expansion needs though adding shelves. Hosts typically connect to the array via FC or iSCSI on the front-end while the array passes the data to disk via FC (older) or SAS (newer) loops. I've not seen a SAN which uses SAS on the front-end for host connectivity.
|
|
#
?
Sep 11, 2014 18:41
|
|
|
NullPtr4Lunch posted:Not that I need any more reasons to hate this machine, but does it really run Windows? Yep, not just that but afaik all Clariion CX and VNX arrays SPs are actually windows machines. The event logs are even in windows format (download the spcollect and open them in event viewer if you want). NippleFloss posted:I'm not sure if it's any different on the VSP, but Hitachi's 9990, which was their to of the line array, team windows for the service processor. Even better, the service processor was really just a Windows laptop tucked away in a slot hidden behind a panel. Yeah the SMU in the VSP is still a windows machine just like the USP/9900/9500 machines. It's just built into the chassis like the later USP's now (it was an XP laptop built onto a blade essentially). zSeries mainframes used to do a similar thing with a thinkpad running OS/2 iirc, though I guess they probably don't do that anymore.
|
|
#
?
Sep 11, 2014 19:01
|
|
|
Docjowles posted:Usually that's for direct attached storage, as opposed to accessing it over the network as you would with the other options. Klenath posted:SAS is used for SAN back-end disk shelf loops and related expansion needs though adding shelves. Hosts typically connect to the array via FC or iSCSI on the front-end while the array passes the data to disk via FC (older) or SAS (newer) loops. I've not seen a SAN which uses SAS on the front-end for host connectivity. Good to know - thanks for the info.
|
|
#
?
Sep 11, 2014 19:56
|
|
|
|
| # ? May 26, 2024 22:43 |
|
|
Klenath posted:SAS is used for SAN back-end disk shelf loops and related expansion needs though adding shelves. Hosts typically connect to the array via FC or iSCSI on the front-end while the array passes the data to disk via FC (older) or SAS (newer) loops. I've not seen a SAN which uses SAS on the front-end for host connectivity. The HP MSA 2000 line is super entry level, comparable to Dell's md3200. I wouldn't be surprised if the SAS ports are for host connectivity (in addition to expansion shelves).
|
|
#
?
Sep 11, 2014 20:00
|
|