Discuss

discuss@lists.centos.org

34753 discussions

Re: [CentOS] Backup server
by Les Mikesell 15 Jan '10

15 Jan '10

Sorin Srbu wrote: > >> So you need to be able to walk the fine line >> between these two. > > I'm trying. Something it just isn't enough. Although the boss has a soft spot > for linux, as he also heads the CADD (Computer Aided Drug Design)-group. > >> To put it into perspective, ask the manager how much it would cost the >> business if this data was unrecoverable? After that, if they still >> don't want to spend a few hundred $$s on the insurance, get it in >> writing that your manager understands the risk and print it out and >> post it on your office wall. > > Rather confrontative isn't it? Me being a Swede, I try to avoid those > situations if possible, and find a compromise instead that both parties can > live with. 8-} Oh, and I'm a government employee, so the money I spend is > tax-payers money. Got to be careful there. Being careful with the money is the point. Someone has to understand the risks. > You know how that saying goes? You can chose between good, fast and cheap. But > you're only ever allowed to pick any two. For me that's IT in a nutshell. ;-) The other question to ask is whether an offsite copy is needed. After a fire or other site disaster some businesses might collect the insurance money and disappear - others might want to be able to rebuild and continue. Government operations would probably need to continue and need a plan for that. -- Les Mikesell lesmikesell(a)gmail.com

2 1

8-15 TB storage: any recommendations?
by Boris Epstein 15 Jan '10

15 Jan '10

Hello everyone, This is not directly related to CentOS but still: we are trying to set up some storage servers to run under Linux - most likely CentOS. The storage volume would be in the range specified: 8-15 TB. Any recommendations as far as hardware? Thanks. Boris.

36 129

Re: [CentOS] Backup server
by Benjamin Franz 15 Jan '10

15 Jan '10

Sorin Srbu wrote: > > Today I have five 500GB-disks raided on linux machine. Remove one for parity > and I have 2TB of real space available. Doing a 0+1, ie 1TB, would indeed be > better as performance goes, but 1TB of space, well, it just isn't enough > unfortunately. > > As it is now, the 2TB shebang is mounted as /backup. Does that count as a > single filesystem? > I was thinking about your long term here. Make sure to use LVM to create your underlaying partition. Then you can add disk space in the future without having to reformat everything and can just grow your ext3/ext4 partition instead. -- Benjamin Franz

2 1

Re: [CentOS] Backup server
by Les Mikesell 15 Jan '10

15 Jan '10

Sorin Srbu wrote: > >>> I've never had any problems with software raid5 in linux before, but you >>> never >>> know... >> There's a big write performance hit from raid5 (software or not). It >> may not be enough to be a showstopper but I wouldn't recommend it. Can >> you reconfigure to a 0+1 or some other type that has better performance >> without losing too much space? The archive does have to be on a single >> filesystem, though, and if you use the epel RPM it makes things easier >> if you mount the volume at /var/lib/BackupPC before the install. >> See 'how to change archive directory' from: >> http://sourceforge.net/apps/mediawiki/backuppc/index.php?title=Tips_and_Tri… >> to do it after the install. > > I know, but in linux it's still a lot better than Windows, which I today > regret I ever introduced it in... Don't get me wrong, reliability is fine, but > the recheck on every restart is kinda' bothersome and takes like forever > > Today I have five 500GB-disks raided on linux machine. Remove one for parity > and I have 2TB of real space available. Doing a 0+1, ie 1TB, would indeed be > better as performance goes, but 1TB of space, well, it just isn't enough > unfortunately. If you have any opportunity to change things, I'd get some larger drives and use raid1 or 0+1. If you want offsite copies, a workable approach in the 2TB scale is to make a 3-member raid1 where you periodically swap the 3rd drive (in an internal or external swappable sata enclosure). If you don't need that, a 4 drive 0+1 raid of 1.5 TB drives would give you 3TB and better performance. What you have will work - but it will limit performance. > As it is now, the 2TB shebang is mounted as /backup. Does that count as a > single filesystem? Yes, but if you use the epel rpm, either mount it at /var/lib/BackupPC or put a symlink there before the install. If you install from the sourceforge source there is an install script that modifies the location so you can put things where you want, but the rpm packages have already done that. The next version will make this easier to change but the current one needs to stay in the location set when the package was built. -- Les Mikesell lesmikesell(a)gmail.com

2 2

Re: [CentOS] Backup server
by nate 15 Jan '10

15 Jan '10

Sorin Srbu wrote: > Already done. Still feel a bit burned by the whole matter though. I still feel burned from IBM's 75GXP fiasco ~7 years ago, even joined the lawsuit at the time(and got booted by the judge because I was in another state), had probably an 80% failure rate on those disks. I got a PDF on a CD somewhere that has all kinds of internal IBM docs(from the lawsuit) showing how they knew what the problems were but refused to fix them. OT but reminded me of that.. nate

4 5

Re: [CentOS] Backup server
by Les Mikesell 15 Jan '10

15 Jan '10

On 1/14/2010 9:08 AM, Sorin Srbu wrote: > >> Go to the vendor's web site, enter their serial numbers and get an RMA for a >> free replacement. Every vendor has had bad batches. > > Already done. Still feel a bit burned by the whole matter though. Unfortunately, the only way commodity priced things get large scale real-world testing is after a large number of them have been sold. If the first one sold had to be perfect none of us could afford to buy it. -- Les Mikesell lesmikesell(a)gmail.com

2 1

Re: [CentOS] Backup server
by Brian Mathis 15 Jan '10

15 Jan '10

On Thu, Jan 14, 2010 at 9:37 AM, Sorin Srbu <sorin.srbu(a)orgfarm.uu.se> wrote: >>-----Original Message----- >>From: centos-bounces(a)centos.org [mailto:centos-bounces@centos.org] On Behalf >>Of Benjamin Franz >>Sent: Thursday, January 14, 2010 2:12 PM >>To: CentOS mailing list >>Subject: Re: [CentOS] Backup server >> >>If you have any budget at all, invest in bigger drives. 7200 RPM 1 TB >>RAID rated drives can be bought for $160 each. Desktop rated 5900 RPM >>1.5 TB drives (which you can probably get away with in a dedicated >>backup server since you don't care a lot about speed and can tolerate >>long pauses for sector repair) can be bought for $110 each. Check Newegg. > > I haven't got a budget really. Today I asked for a new group-printer today and > the boss looked pained... 8-} > > I opted for the proven 500GB-sized disks and got more of those instead. I've > had a handful of 750GB-drives die on me recently. Somehow it feels the > technology isn't quite there yet for the bigger drive-sizes. Anybody remember > the IBM Deskstars in the early 00's...? > > Also, my experience is the more smaller disks you have, the faster they get. > Less to write to each I guess. > > >>Second, to maximize 'depth' of backups you should use a 'Tower of >>Hanoi'-like backup system. > > Good advice, thanks! > -- > /Sorin There seems to be a persistent conception among managers that anything "IT related" is a huge capital expenditure (as it used to be), and there's all sorts of resistance to buying anything new. However, you probably spend more on printer paper in 1 week than it costs to buy a 1TB drive. This kind of equipment is a disposable commodity, even though the accounting department still prefers to write it off over 7 years. However, IT also has a reputation of always wanting to buy new toys. Many times these toys are not needed, even though the IT person insists that they are. So you need to be able to walk the fine line between these two. To put it into perspective, ask the manager how much it would cost the business if this data was unrecoverable? After that, if they still don't want to spend a few hundred $$s on the insurance, get it in writing that your manager understands the risk and print it out and post it on your office wall.

5 6

Are SSD disks worth the cost for server usage?
by Mathieu Baudier 15 Jan '10

15 Jan '10

Hi, we want to upgrade our infrastructure and are considering renting two different types of server by a provider: - A: one is with 80 GB SSD (and 12 GB memory) http://www.ovh.co.uk/products/eg_ssd.xml - B: the other with 750 GB SATA2 (and 8 GB memory). http://www.ovh.co.uk/products/eg_best_of.xml The target infrastructure will host: - a few public web sites (PHP / MySQL) - a development infrastructure (subversion, bugzilla, automated builds) This will use KVM virtualization with Cent OS 5.4 hosts and guests. On the long run we will probably rent quite a few of type B depending on our needs of space and computing power. My question is whether it is worth using one of type A (with SSD drives) in order to host the critical data such as the web sites and the subversion repository (critical in the sense that these services should be fast, always available, and the related data as safe as possible). This critical server will be more read oriented than write. We will probably use a hardware RAID 1 with battery backed unit on it. My understanding is that SSD are much much faster. Is it really true in the real world? Are SSD drives also safer? (that is, less likely to crash under load) Or to the contrary? I have read/heard conflicting opinions so far and I'll be happy to hear opinions/experiences on SSD drives in a server settings from people on the list. More general comments about our plans are always welcome of course! Cheers, Mathieu PS: we will actually rent them in France, but I put links to the equivalent in Britain in order to have the description in English. For reference and for the French-speaking people, here are the actual servers we consider: http://www.ovh.com/fr/produits/eg_ssd.xml http://www.ovh.com/fr/produits/eg_best_of.xml (they are exactly the same)

7 10

[story] Thank goodness for links and caching DNS
by Michael A. Peters 15 Jan '10

15 Jan '10

Right in the middle of doing something important on my Ubuntu box, the web quit working. However, I could get to sites in my /etc/hosts file. Sure enough, all three of my ISPs nameservers were down. I'm not a DNS guy, but on my CentOS boxes I always installed a caching DNS out of the box since it was part of the obvious install options. So I set the DNS to one in my Ubuntu box, and of course the port was closed. So I ssh'd on in, opened 53 in the firewall, and the port was still closed. Tried running Firefox from the CentOS box via X forwarding over ssh but oddly, that seemed to lauch local Ubuntu firefox which didn't work. All of my CentOS boxes are headless, so all my boxes with working nameserver resolution thus only had text internet, which I haven't used in years. locate named.conf only showed some file in dbus that I knew wasn't right, but I was able to use links and find what I needed on google, the caching nameserver conf file has a slightly different filename than standard bind named.conf. I followed the warning, installed the system config utility for modifying it, but damn - for someone who doesn't know bind, that tool is scary looking, intimidating, I was afraid I was going to break my working caching nameserver if I messed around in it. So I backed up the file and hand edited it to add the IP to listen on, restarted bind, and am back in business. This is the second time in the last 6 months that all three of my ISP's nameservers have gone down, I wonder if they really are the same physical box, maybe even same network interface. What's the point of them having three if they all go down at same time? Ah well, I don't need them anymore for that. Thankfully the caching nameservers were already installed on my CentOS boxes, two of them now listens on their external interface, never again will I use my ISPs nameserver.

2 2

High load since passing from rhas3 to centos4.8
by fortin.pierre＠bell.ca 15 Jan '10

15 Jan '10

We are migrating HP servers from RHAS3 to centos 4.8. Since, the load average is reaching really high values. Under usage we see loads of 15-20 instead of the 0.5-1.5 we were used to. I searched a lot for information on such an issue. Iostat, vmstat, top, ps. But I don't get significant hint. Cpu usage remains low, IO wait remains low, no disks are lagging, no swap usage. Are there some tool / other way to diagnose why the load average is high? Like which processes are waiting, where they are stuck. Are there calls to drivers or system process that are slow? Thanks

5 7

Jump to page:

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Discuss