Re: Linux, gobs of RAM, RAID and performance suckage...
On Nov 30, 2006, at 11:59, Paul Lussier wrote: However, now when backups are run, the system becomes completely unresponsive from an NFS client perspective, and the load average skyrockets (e.g. into the 40s!). Does anyone have any ideas ? I'm at a complete loss on this one. Have you tried taking the RAM out and seeing if performance reverts? If so... are the RAM sticks the correct type, properly installed for memory interleaving, etc? Did the BIOS decide to reset itself when new RAM was installed? Could the BIOS be buggy with 4GB of RAM (that thing with the upper 640MB of RAM being masked)? When you're getting desperate you can reset the BIOS to known-good/failsafe/optimized defaults and see if anything changes. I've forgotten some 2.4 stuff but there was a big-mem version of the 2.4 kernel at one point to work around problems with too much RAM. All guesses... -Bill - Bill McGonigle, Owner Work: 603.448.4440 BFC Computing, LLC Home: 603.448.1668 [EMAIL PROTECTED] Cell: 603.252.2606 http://www.bfccomputing.com/Page: 603.442.1833 Blog: http://blog.bfccomputing.com/ VCard: http://bfccomputing.com/vcard/bill.vcf For fastest support contact, please follow: http://bfccomputing.com/support_contact.html ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
Bill McGonigle writes: I've forgotten some 2.4 stuff but there was a big-mem version of the 2.4 kernel at one point to work around problems with too much RAM. Ah, that's right. If you're going to run 2.4 with 1GB RAM you need to apply the rmap patches or performance will get worse the more RAM you have. Severely so with 2GB. Google says rmap patches are here, but the server is giving me a 403 at the moment. http://www.surriel.com/patches/ -- Dave ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
On Thursday 30 November 2006 11:59 am, Paul Lussier wrote: Before yesterday we were noticing lots of NFS drop-outs on the clients (300+ of them) and we correllated this pretty much to the backups (amanda). The theory was that local disk I/O was beating out nfs-client requests. I'm not sure the topology of your SAN, but can you connect another machine to the SAN with read-only access to those filesystems to do backups without involving the NFS server at all? -N ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
On 11/30/06, Paul Lussier [EMAIL PROTECTED] wrote: This is bizarre. Spare memory will ALWAYS be used to cache. This is fine and 'normal'. -- -- Thomas ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
Neil Joseph Schelly [EMAIL PROTECTED] writes: I'm not sure the topology of your SAN, It's not a SAN. It's direct-attached storage. but can you connect another machine to the SAN with read-only access to those filesystems to do backups without involving the NFS server at all? -N In theory, yes, in practicality, no. Also, the backups are not touching the NFS server. Cast of characters: Playing the part of:Actor: the NFS server space-monster the backup server amanda Amanda connects to the amanda daemon on space-monster (via inetd) and requests that he start his backup process. space-monster in turn kicks off a gtar process which sends the data back to amanda. The gtar process is reading the local disk, not NFS. Apparently, this local disk I/O request supercedes NFS disk requests. But there's much more than that happening here. CPU utilization is through the roof on space-monster. That wasn't the case yesterday before the memory upgrade. -- Seeya, Paul -- Key fingerprint = 1660 FECC 5D21 D286 F853 E808 BB07 9239 53F1 28EE A: Yes. Q: Are you sure? A: Because it reverses the logical flow of conversation. Q: Why is top posting annoying in email? ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
Thomas Charron [EMAIL PROTECTED] writes: On 11/30/06, Paul Lussier [EMAIL PROTECTED] wrote: This is bizarre. Spare memory will ALWAYS be used to cache. This is fine and 'normal'. I was not implying that the use of spare memory as cache was bizarre. I *know* that spare memory will be used this way, and that in general this is a good thing. What is bizaarre is that one could *INCREASE* system resources and performance *GETS WORSE*. This is unintuitive and inverse to most normal expectations, IOW, BIZARRE! -- Seeya, Paul -- Key fingerprint = 1660 FECC 5D21 D286 F853 E808 BB07 9239 53F1 28EE A: Yes. Q: Are you sure? A: Because it reverses the logical flow of conversation. Q: Why is top posting annoying in email? ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
Neil Joseph Schelly wrote: On Thursday 30 November 2006 11:59 am, Paul Lussier wrote: Before yesterday we were noticing lots of NFS drop-outs on the clients (300+ of them) and we correllated this pretty much to the backups (amanda). The theory was that local disk I/O was beating out nfs-client requests. Its been years since I've been in the guts of NFS. Things I remember for server tuning are: * Make sure the user-mode processes aren't running into ulimit problems. * Lockd chews up a lot of kernel resources - just which ones depends on its implementation. Check shared memory and semaphores for resource depletion. * NFS over TCP is a real system hog. * The socket queue size (which used to be a kernel configurable option, I don't about today's kernels), if set larger would improve performance. * MTU should be uniform across the network. Things will work if its not uniform, but performance will suffer. In general, bigger is better - up to the maximum size the servers/clients can handle. And old problem was that connections would linger after they've been closed. I think the modern kernels/NFS distributions have fixed that though. The fastest way to clog NFS is with a bunch of small random access read-writes. This is typically manifested in directory updates (for example: updating a netnews site over NFS), trying to run a database across an NFS connection with a lot of simultaneous writers, and having too many umbrellas in your empty rum glass. Otherwise, if you can wait until I'm able to dig up my old notes on NFS load-testing, I can get some more ideas. Right now, I'm several thousand miles from my filing cabinet. --Bruce ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
On Thursday 30 November 2006 01:51 pm, Paul Lussier wrote: It's not a SAN. It's direct-attached storage. Winchester OpenSAN FC-based RAID array Isn't this the storage? I assume the description meant it was a SAN. What is the topology of that FC network? In theory, yes, in practicality, no. Also, the backups are not touching the NFS server. Amanda connects to the amanda daemon on space-monster (via inetd) and requests that he start his backup process. space-monster in turn kicks off a gtar process which sends the data back to amanda. The gtar process is reading the local disk, not NFS. Apparently, this local disk I/O request supercedes NFS disk requests. But there's much more than that happening here. CPU utilization is through the roof on space-monster. That wasn't the case yesterday before the memory upgrade. It sounds like the NFS server, space-monster in your description, is having to read all that data from the filesystem and send all that data to the backup server called amanda in your description. If you could put another server on the SAN, in the FC network, that machine could also mount the partition (read-only obviously) on the SAN and backup the contents without the space-monster even knowing anything is happening. -N ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
Paul Lussier writes: This is bizarre. We've got an NFS server with Dual 3Ghz Xeon CPUs as our NFS server connected to a Winchester OpenSAN FC-based RAID array. The array is a single 1TB partition (unfortunately). Before yesterday we were noticing lots of NFS drop-outs on the clients (300+ of them) and we correllated this pretty much to the backups (amanda). The theory was that local disk I/O was beating out nfs-client requests. We also noticed that our memory utilization was through the roof. We had 2GB of PC2300, ECC, DDR, Registered RAM. That RAM was averaging the following usage patterns: active - 532M inactive - 1.2G unused - 39M cache - 1.3G slab cache - 255M swap cache - 6M apps - 78M buffers- 350M swap - 11M We were topping out our memory usage and occasionally dipping into swap space on disk. Yesterday we added 2GB of RAM and our memory utilization now looks like this: active - 793M inactive - 2.3G unused - 213M cache - 2.9G slab cache - 194M swap cache - 2M apps - 71M buffers- 313M swap - 4.5M So, it appears we really only succeeded in doubling the cache available to the system, and a little more than halving the amount of swap that was getting touched. However, now when backups are run, the system becomes completely unresponsive from an NFS client perspective, and the load average skyrockets (e.g. into the 40s!). Does anyone have any ideas ? I'm at a complete loss on this one. General nfs server comments: 1) Make sure you are exporting the nfs shares async otherwise most operations will seem slow from the clients point of view. Check /proc/fs/nfs/exports for 'async' If not there, set it in your /etc/exports 2) Make sure you server has enough nfsd threads to handle the client load. With 300 clients, you should have at least 100 nfsd threads in my opinion. Check your init.d scripts for how to set this. Other stuff: 11MB of swap doesn't mean anything is wrong. It's actually a good thing meaning your kernel found 11MB of stuff that wasn't needed and booted it out to swap space in order to make more room available for cache. You should check the disk rates and cpu usage with 'vmstat 5' for a few minutes. This will also show how much of the cpus are spending in wait. The full outout /proc/meminfo may also be useful. I assume you are using a x86_64 kernel? Using a 32bit kernel is ok as long as you dont run out of low memory. The kernel's heap/slab as well as filesystem metadata (buffers) must come from low memory while filesystem data (cache) and userspace processes can be in high memory. If your SAN is on a dedicated lan to the server only you should investigate converting that subnet to support jumbo frames. Since your using Xeons not Opterons you should make sure irqbalance is installed and running to spread irq load across all cpu cores (this may not be a good idea on a multi-node Opteron system though) You can run top in multi-cpu mode (press '1') to see if any cpus are overloaded with irq or wait load while others are idle. -- Dave ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
Drew Van Zandt [EMAIL PROTECTED] writes: Can you ask amanda to do bandwidth limiting? I'd think this would be a standard feature... Yes, at least with respect to network bandwidth. I'm not sure about Disk I/O bandwidth. But I don't think it's an amanda thing, I think it's a poorly tuned kernel thing. Possibly combined with an ancient buggy kernel thing. And not just little of a Linux makes a crappy NFS server thing :) -- Seeya, Paul -- Key fingerprint = 1660 FECC 5D21 D286 F853 E808 BB07 9239 53F1 28EE A: Yes. Q: Are you sure? A: Because it reverses the logical flow of conversation. Q: Why is top posting annoying in email? ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
Paul Lussier writes: Yesterday we added 2GB of RAM and our memory utilization now looks like this: active - 793M inactive - 2.3G unused - 213M cache - 2.9G slab cache - 194M swap cache - 2M apps - 71M buffers- 313M swap - 4.5M When you are in this condition, how large is Dirty and Writeback in /proc/meminfo? If you have a large amount of Dirty (read: 300MB) and the underlying filesystem is using ordered journaling (the default for ext3) you can cause very long delays if a process requests a fsync on a file (vi does this alot). Instead of writing a single file to disk right away the filesystem must write all other outstanding data to disk first! If your clients are write heavy and you are using ext3 you should consider data=writeback when mounting the filesystem. I've run into this issue on a server with 64GB of RAM with dirty commonly exceeding 4GB under load. Poor users of vi had to keep waiting for the server to write out 4GB of data every time they saved a file. data=writeback helped tremendously as only the previous metadata is forced to disk on a fsync() instead of everything. Correct. The nfsd's on the server get pushed to the bottom of the queue (in top) and what I see is [kupdated] and tar (from the backups) rotating between the top position. There was one or two other processes up there too, one of which I think was a kernel thread ([bdflush] ?). As far a I know, the tar's were not making processes, but I can't be totally sure of that. Nothing *else* was making progress, that's for sure. kupdated?!?! A 2.4 kernel??!?!? -- Dave ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
Re: Linux, gobs of RAM, RAID and performance suckage...
Dave Johnson [EMAIL PROTECTED] writes: kupdated?!?! A 2.4 kernel??!?!? Ahm, yeah. Are you shocked at kupdated running with a 2.4 kernel because it shouldn't be there with a 2.4 kernel, or shocked I'm still running a 2.4 kernel? -- Seeya, Paul -- Key fingerprint = 1660 FECC 5D21 D286 F853 E808 BB07 9239 53F1 28EE A: Yes. Q: Are you sure? A: Because it reverses the logical flow of conversation. Q: Why is top posting annoying in email? ___ gnhlug-discuss mailing list gnhlug-discuss@mail.gnhlug.org http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/