@ChrisS It's a box for streaming. Need to maximise throughput and the official recomendation is to leave them seperate and allow the software to deal with balancing the load.
Any kind of teaming in the middle would just slow stuff down
We normally keep it all within one chassis, but this is an upgrade project using some of their existing kit so unfortunately it's external to the chassis where the hypervisors are. So there's an added layer of comlpexity
Perl is probably the most commonly abused DevOps language; but Ruby is more likely to get your an interview. JavaScript may be on the horizon, with both Google and Joyent seeming to think so...
I have a VMWare VM and data set that compressed between 70%-80% on ZFS. My client has a NetApp, though... not sure which model. But they updated to OnTap 8.x last week... and apparently, compression was available.
@basil, I'm reading some Netapp support notes from the client attached to an email chain...
When you updated the firmware, the module was taken completely offline, and since this is a single path system, the loss of the module causes the shelf to disappear.
Had the filer been a multipath HA system, the shelf would have offlined the A module and switched to the B module on a secondary loop to continue serving data. The filer would not have lost access to the shelf. Because the system is single path, from the filer’s point of view the shelf simply vanished.
They apparently blew up their Netapp in a firmware update...
@ewwhite Any storage with a single point of failure should be considered unusable for prod. Storage touches everything the business does, and can't ever go down
@Basil I can't control every client. A lot of places have an internal staff... and those staff don't have much to do. The prospect of a SAN or VMWare project excites them, and they usually don't listen to me.
I'm talking with the place that bought into a VNXe next week... and my input was not taken at all... Granted, I don't know storage :)
@ColdT Looks like it - basically, with only one binding the DHCP clients picks up the address within 2 seconds. Two bindings mostly works, but normally takes about 3 or 4 seconds to get the address. With 3 bindings, it takes a minimum of 4 seconds, sometimes longer and it often fails altogether
@Basil I don't think they KNOW the difference. I think once they start paying $40k+ and are told it's a "SAN", they expect it to be more than a box of disks.
Maybe the storage vendors not giving the right advice.
Anyone who tells you they don't have money for storage resiliency is lying. What they're actually saying is that they don't value business continuity as much as they they value whatever else the money will be spent on. The reality is that 40k on storage that will fail hard on a bad day is more expensive than 60k for something that will be stable
storage sales are just plain aggressive, they want their kit brought, they will go through hell to convince you, and unless you know something about storage, you'll fall into the trap and be brought
I'm intervening when I can... I convinced another place to move their Equallogic quote from an all 7200rpm setup to 10k 2.5's. Still no SSD's... but it's coming from the 14-disk RAID 5 on SATA in their current unit.
@ColdT My point is that if you're running critical applications, it doesn't matter what size business you run. Can an SMB trucking company run trucks with 17 wheels?
@ewwhite yeah, and considering that (as far as I know) that's only 2000$ or so...
@ColdT You never need anything, but if you're going to buy storage and you're going to run all your servers on it, you're better off spending the money you would have spent on redundant power supplies on eliminating storage single points of failure
@ColdT bullshit that you have to do that though - we got a quote for a content filtering system. it started out around 20k which we didn't go for. after about a month of this company trying to sell it to us, they brought it down to 7k - why did you not give me that price in the first place
netapp aren't that bad, i find them easier to manage compared to emc units
@lsiunsuex same, we had riverbed do that to us, was quote single unit 20k'ish and was brought down to 10k for two units (which of course we was over the moon with lol)
AND if i goto my boss with a quote for 20k, he calls them and gets it down to 15k - how is that gonna make me look infront of my boss? you think i'm gonna want to deal with you anymore now that you made me look like shit in front of him ?
Had a quote for home insurance a few weeks ago which I didn't take. Had a guy call up to tell me how he can help reduce the quote - told him he can forget it if he can't be arsed to give me the best price int he first place. It ended in somewhat of a row :D
I am going to start a migration from and Exchange 2003 cluster to a new Exchange 2010 SP2 with 14 mailbox databases of a supposed maximum size of about 400 GB each. Each database will be divided into 2 LUNS, data and transaction logs. LUN size takes into account to leave about 20% of free space t...
I dunno. The way I read it, it's pretty clear that there are 14 mailbox databases across 28 LUNs for data/logs and each database should have a max size of 400GB
I have two servers Main and Backup, such that I want a Backup server to be able
to replace the Main one as fast as possible without physically accessing them.
So I want the Main and the Backup server to see same hard drives. (Even if backup does not mount them). Also I don't want stuff like NF...
@MDMarra You're (sort of) wrong actually. HP makes (made?) a single "clustered" chassis that had two nodes in it that could see the same set of hard drives as well as their own individual ones.
So… when the corrected ECC error count on my hard drives hits a billion, is that when I should start to get worried?
Realistically, you don't/shouldn't share DAS. You can buy a 3rd server and setup a fairly inexpensive system to share the storage on the 3rd server (Starwind/FreeNAS/etc) — Rex1 min ago
^ Is this true?
I thought that with some DAS, you can absolutely present LUNs to multiple servers, and as long as they play nice with clustered filesystems, etc, it would work fine.
@MikeyB That just made me burst out laughing in the middle of the office while talking to some MS "technician" who told me I had to ring a different number to find out how to buy their software (I had to put them on hold to calm down). My favourite was: 919 sad
@Roomey You mean w.r.t. the .bash_history excerpt above? We told the customer they should stop using volunteer student "sysadmins" from the university.
Hahahaha it happens again. Co-worker: "Oh boy, $customer's admin really screwed up." Me: "What did he do this time?" C: "chmod 755 -R /" "A week ago. Backups haven't run since."
I hate when people ask questions where you think they're in over their head, but you're not sure, so you really don't want to hand them a gun pointed at their foot, but you don't want to call them a noob either.
My 403b has been fucked up at my new job, so there haven't been any company contributions since I was hired. Just today, 3 months worth of 15% contributions were added all at once
@MikeyB That was very calm of you. I don't know how many times one of our (ex)-admins has left the root password in the bash history file, but after seeing your experience I'll never complain again!
@ewwhite I would have thought, if a human has time to see what is happening and hit the button, it is probably already too late! how much is traded per second usually?
tar -cvf test CN_Rail_SANReports.zip CN_Rail_SANReports.zip CN_Rail_SANReports.zip a CN_Rail_SANReports.zip 94745 blocks. a CN_Rail_SANReports.zip 94745 blocks. a CN_Rail_SANReports.zip 94745 blocks.
@MikeyB Any file will do, but now that I have this one, I'll use it
I am looking for a solution which allows me to access any file of my productive environment on my local virtual machine. Both systems are running Ubuntu 10.04 lts.
The current solution is a rsync however this means that we have to transfer every file which might be useful. Hovewever this costs a...
It sounded to me like he wants a system where he can develop against a copy of his production files, but he doesn't want to have to download all of them to his dev machine.
He sounds like he's looking for something that can download them to his VM as they are needed by his development process
Either way, it's a shitty question with no real details
You can easily use python to generate such a tarfile:
mktar.py:
#!/usr/bin/python
import datetime
import sys
import tarfile
tar = tarfile.open(fileobj=sys.stdout, mode="w|")
info = tarfile.TarInfo(name="fizzbuzz.data")
info.mode = 0644
info.size = 1048576 * 16
info.mtime = int(datetime.datet...
I have a 24 core machine with 94.6GiB RAM running Ubuntu server 10.04. The box is experiencing high %iowait, unlike another server we have (4 cores) running the same types and amounts of processes. Both machines are connected to a VNX Raid fileserver, the 24-core machine via 4 FC cards, and the o...
@ewwhite "Ignore above about dm and sdc, they are all internal disks, and not part of the problem." So he's posted no actual useful data. Is he using NFS mounts? Are sda and sdb the LUNs?
Can anyone figure out what the fuck this guy is talking about
Yes, I mean external DAS. But I need it to be (a) as efficient as internal one physically sitting in the server - ZERO overhead; (b) it should be accessible from different server as active-passive. i.e. Backup would mount the volume if-and-only-if Main is down. I'm not looking for a cable running inside other server and using its disks — Artyom1 hour ago
With puppet, how might have multiple sections in a single file with templating? For example, working on how I want to go about making one for haproxy. A single config has multiple frontends and backends, and each frontend and backend will have it's own set of variables
It is like I need to itterate over an array of objects, and the object will have multiple attributes
The objects would be "backend A", "Backend B". And each backend would have properties like "option foo" and then maybe even arrays within that
Him: «Rescan command not working [root@db ~]# ./rescan-scsi-bus.sh -i -bash: ./rescan-scsi-bus.sh: No such file or directory» Me: "Install sg3_utils for the rescan-scsi-bus.sh command" Him: "But I still dint re-scan yet , do I need to install sg3_utils for this" Me: Derp?
I am looking for a solution which allows me to access any file of my productive environment in my local virtual machine for testing. Both systems are running Ubuntu 10.04 lts.
The current solution is a combination of cron and rsync.
This means that we have to transfer periodically every producti...
I have two servers Main and Backup, such that I want a Backup server to be able
to replace the Main one as fast as possible without physically accessing them.
So I want the Main and the Backup server to see same hard drives. (Even if backup does not mount them). Also I don't want stuff like NF...
You can easily use python to generate such a tarfile:
mktar.py:
#!/usr/bin/python
import datetime
import sys
import tarfile
tar = tarfile.open(fileobj=sys.stdout, mode="w|")
info = tarfile.TarInfo(name="fizzbuzz.data")
info.mode = 0644
info.size = 1048576 * 16
info.mtime = int(datetime.datet...
