I'm having a server report a high load average when backing up Postgres database files to an external USB drive. This is driving my loadbalancers all out of kilter and causing a large volume of network monitor alerts.
I have a 1TB USB drive plugged into a USB2 port that I use to back up the production drives (which are SCSI). It's working fine, but while doing backups (hourly) the load average on the server shoots up from the normal 0.5 - 1.5 or so up to a high between 10 and 30. Strangely, even though the "load is high" the server is completely responsive, even the USB drives being accessed are!
Backup script is really simple, run via cron, pretty much just:
#! /bin/sh hour=`date +%k`; pg_dump <options> mydatabase > /media/backups/mydatabase.$hour.pgsql;
where /media/backups is the mount point for the USB drive.
Using top to diagnose, nothing seems to be particularly high! IoWait seems reasonable (10-30%) and CPUs are 0.5%, Idle is 70-90%. Even accessing the USB partition while the load is "high" is responsive!
I'm guessing that something changed in how load average is counted?
Server Stats: Late model 8-way Xeon, SuperMicro brand. CentOS 4.x / 64 (all updates applied, booted after last kernel update) Kernel 2.6.9-89.0.16.ELsmp 4 GB ECC RAM 300 GB SCSI HDD. Standard Apache/PHP, Postgres 8.4.
Any idea how to revert to the old load average tracking behavior short of using a stale and potentially insecure kernel?
Benjamin Smith wrote, On 11/16/2009 10:56 PM:
I have a 1TB USB drive plugged into a USB2 port that I use to back up the production drives (which are SCSI). It's working fine, but while doing backups (hourly) the load average on the server shoots up from the normal 0.5 - 1.5 or so up to a high between 10 and 30. Strangely, even though the "load is high" the server is completely responsive, even the USB drives being accessed are!
Backup script is really simple, run via cron, pretty much just:
#! /bin/sh hour=`date +%k`; pg_dump <options> mydatabase > /media/backups/mydatabase.$hour.pgsql;
where /media/backups is the mount point for the USB drive.
Using top to diagnose, nothing seems to be particularly high! IoWait seems reasonable (10-30%) and CPUs are 0.5%, Idle is 70-90%. Even accessing the USB partition while the load is "high" is responsive!
I'm guessing that something changed in how load average is counted?
Server Stats: Late model 8-way Xeon, SuperMicro brand. CentOS 4.x / 64 (all updates applied, booted after last kernel update) Kernel 2.6.9-89.0.16.ELsmp 4 GB ECC RAM 300 GB SCSI HDD. Standard Apache/PHP, Postgres 8.4.
Any idea how to revert to the old load average tracking behavior short of using a stale and potentially insecure kernel?
Note, although I have a couple of ideas, I am answering/questioning more out of curiosity than experience. salt appropriately.
Are you saying that when you were running a previous kernel the same operations with the same devices did not have the high load? Which specific kernels worked as desired (if someone is going to bisect the problem they need a start point)?
Are there other processes on the machine that are waiting to use the db while the dump is occurring? How many postgres processes are waiting for the dump to finish (it has been a while since I ran postgres so I don't recall how it deals with query's during a dump)?
As workarounds perhaps asking the kernel to schedule in a specific way might help, i.e.: #1 set the backup on a particular set of processors, # replace the pg_dump line above with taskset -c 3-4 pg_dump <options> mydatabase > \ /media/backups/mydatabase.$hour.pgsql;
#2 set the usb-storage on a particular set of processors, # Note USBSTORPID= line prototyped on CentOS 5 machine not 4. USBSTORPID=`ps aux |grep usb-storage|head -1 |awk '{print $2}'` taskset -p -c 3-4 $USBSTORPID #you might even go back and reduce the processor list #to just 3 or 4 instead of both.
#3 don't update atime # (should at worst be a minor thing, and you say that # the usb mounted file system is responsive, # but perhaps it would help some.) mount -oremount,noatime /media/backups/
I have not had the taskset of the USB driver cause faults when used on a dual processor Xeon, but if any of the above breaks your system you get to keep the chunky bits. :0
See comments below...
On Tuesday 17 November 2009 07:52:01 Todd Denniston wrote:
Benjamin Smith wrote, On 11/16/2009 10:56 PM:
I have a 1TB USB drive plugged into a USB2 port that I use to back up the production drives (which are SCSI). It's working fine, but while doing backups (hourly) the load average on the server shoots up from the normal 0.5 - 1.5 or so up to a high between 10 and 30. Strangely, even though the "load is high" the server is completely responsive, even the USB drives being accessed are!
Backup script is really simple, run via cron, pretty much just:
#! /bin/sh hour=`date +%k`; pg_dump <options> mydatabase > /media/backups/mydatabase.$hour.pgsql;
where /media/backups is the mount point for the USB drive.
Using top to diagnose, nothing seems to be particularly high! IoWait seems reasonable (10-30%) and CPUs are 0.5%, Idle is 70-90%. Even accessing the USB partition while the load is "high" is responsive!
I'm guessing that something changed in how load average is counted?
Server Stats: Late model 8-way Xeon, SuperMicro brand. CentOS 4.x / 64 (all updates applied, booted after last kernel update) Kernel 2.6.9-89.0.16.ELsmp 4 GB ECC RAM 300 GB SCSI HDD. Standard Apache/PHP, Postgres 8.4.
Any idea how to revert to the old load average tracking behavior short of using a stale and potentially insecure kernel?
Are you saying that when you were running a previous kernel the same operations with the same devices did not have the high load?
Correct!
Which specific kernels worked as desired (if someone is going to bisect the problem they need a start point)?
kernel-smp-devel-2.6.9-89.0.15.EL (I always keep my machines updated on at least a weekly scheduule)
Are there other processes on the machine that are waiting to use the db while the dump is occurring?
No. Database is actually on a different machine and backups are being done over the network.
How many postgres processes are waiting for the dump to finish (it has been a while since I ran postgres so I don't recall how it deals with query's during a dump)?
One - the one performing the backup. Postgres uses MVCC so pg_dump doesn't block any other connections from continuing/finishing.
As workarounds perhaps asking the kernel to schedule in a specific way might help, i.e.: #1 set the backup on a particular set of processors, # replace the pg_dump line above with taskset -c 3-4 pg_dump <options> mydatabase > \ /media/backups/mydatabase.$hour.pgsql;
There are 8 cores on the machine, none of which are reporting more than 5% load. That's what has me perplexed. When I run top, I see a max of about 30% user. Everything else is zero. When I run the backup script to a non-USB drive, the load average is completely normal (below 0.50, often below 0.10)
#2 set the usb-storage on a particular set of processors, # Note USBSTORPID= line prototyped on CentOS 5 machine not 4. USBSTORPID=`ps aux |grep usb-storage|head -1 |awk '{print $2}'` taskset -p -c 3-4 $USBSTORPID #you might even go back and reduce the processor list #to just 3 or 4 instead of both.
Could you explain to me what this should accomplish? I'm curious as to why you went this route...
#3 don't update atime # (should at worst be a minor thing, and you say that # the usb mounted file system is responsive, # but perhaps it would help some.) mount -oremount,noatime /media/backups/
Already mounted noatime... here's the mount line in the backup script: # mount -o rw,noatime -t ext3 /dev/sdc1 /home/backup/localdb/
Benjamin Smith wrote, On 11/17/2009 01:46 PM:
See comments below...
On Tuesday 17 November 2009 07:52:01 Todd Denniston wrote:
Benjamin Smith wrote, On 11/16/2009 10:56 PM:
I have a 1TB USB drive plugged into a USB2 port that I use to back up the production drives (which are SCSI). It's working fine, but while doing backups (hourly) the load average on the server shoots up from the normal 0.5 - 1.5 or so up to a high between 10 and 30. Strangely, even though the "load is high" the server is completely responsive, even the USB drives being accessed are!
Using top to diagnose, nothing seems to be particularly high! IoWait seems reasonable (10-30%) and CPUs are 0.5%, Idle is 70-90%. Even accessing the USB partition while the load is "high" is responsive!
you might add another field to top while you are watching, Last used cpu (SMP), i.e., start top press f press j press enter
this should let you see if your process is bouncing between processors.
As workarounds perhaps asking the kernel to schedule in a specific way might help, i.e.: #1 set the backup on a particular set of processors, # replace the pg_dump line above with taskset -c 3-4 pg_dump <options> mydatabase > \ /media/backups/mydatabase.$hour.pgsql;
There are 8 cores on the machine, none of which are reporting more than 5% load. That's what has me perplexed. When I run top, I see a max of about 30% user. Everything else is zero. When I run the backup script to a non-USB drive, the load average is completely normal (below 0.50, often below 0.10)
USB chewing up more CPU than normal disks has been my experience all along, this just seems a little extreme.
#2 set the usb-storage on a particular set of processors, # Note USBSTORPID= line prototyped on CentOS 5 machine not 4. USBSTORPID=`ps aux |grep usb-storage|head -1 |awk '{print $2}'` taskset -p -c 3-4 $USBSTORPID #you might even go back and reduce the processor list #to just 3 or 4 instead of both.
Could you explain to me what this should accomplish? I'm curious as to why you went this route...
Even though the process is not using much processor time, having it bounce around between processors can: * thrash the cache of each processor as it goes there * waste time context switching in the next processor * bounce other processes around and cascade the same effects as they go along
I know that there has been some scheduler work over time to have these switches be less likely, but I have also seen some good effects by locking certain processes into a processor instead of letting it float. Usually the best processes to do to are ones that use large amounts of memory, like X or Firefox which are large enough that they thoroughly toss anything else out of a processor's cache.
On Tuesday 17 November 2009 15:37:24 Todd Denniston wrote:
Benjamin Smith wrote, On 11/17/2009 01:46 PM:
See comments below...
On Tuesday 17 November 2009 07:52:01 Todd Denniston wrote:
Benjamin Smith wrote, On 11/16/2009 10:56 PM:
I have a 1TB USB drive plugged into a USB2 port that I use to back up the production drives (which are SCSI). It's working fine, but while doing backups (hourly) the load average on the server shoots up from the normal 0.5 - 1.5 or so up to a high between 10 and 30. Strangely, even though the "load is high" the server is completely responsive, even the USB drives being accessed are!
Using top to diagnose, nothing seems to be particularly high! IoWait seems reasonable (10-30%) and CPUs are 0.5%, Idle is 70-90%. Even accessing the USB partition while the load is "high" is responsive!
you might add another field to top while you are watching, Last used cpu (SMP), i.e., start top press f press j press enter
this should let you see if your process is bouncing between processors.
The process pg_dump is "adhering" fine to processor 1. I see usb-storage bouncing between processors - I've seen it on 3, 4, 7 over perhaps a minute. What could you recommend next?
Benjamin Smith wrote, On 11/18/2009 06:11 PM:
On Tuesday 17 November 2009 15:37:24 Todd Denniston wrote:
Benjamin Smith wrote, On 11/17/2009 01:46 PM:
See comments below...
On Tuesday 17 November 2009 07:52:01 Todd Denniston wrote:
Benjamin Smith wrote, On 11/16/2009 10:56 PM:
I have a 1TB USB drive plugged into a USB2 port that I use to back up the production drives (which are SCSI). It's working fine, but while doing backups (hourly) the load average on the server shoots up from the normal 0.5 - 1.5 or so up to a high between 10 and 30. Strangely, even though the "load is high" the server is completely responsive, even the USB drives being accessed are!
Using top to diagnose, nothing seems to be particularly high! IoWait seems reasonable (10-30%) and CPUs are 0.5%, Idle is 70-90%. Even accessing the USB partition while the load is "high" is responsive!
you might add another field to top while you are watching, Last used cpu (SMP), i.e., start top press f press j press enter
this should let you see if your process is bouncing between processors.
The process pg_dump is "adhering" fine to processor 1. I see usb-storage bouncing between processors - I've seen it on 3, 4, 7 over perhaps a minute. What could you recommend next?
try #2 set the usb-storage on a particular set of processors, # Note USBSTORPID= line prototyped on CentOS 5 machine not 4. USBSTORPID=`ps aux |grep usb-storage|head -1 |awk '{print $2}'` taskset -p -c 4 $USBSTORPID
and still I have not had the taskset of the USB driver cause faults when used on a dual processor Xeon, but if any of the above breaks your system you get to keep the chunky bits. :0
so if you try it, keep an eye on it. reversing the above taskset in your case would I _think_ be: taskset -p -c 0-7 $USBSTORPID
Sorry can't suggest much about the usb issue but for such frequent backups, as well as to enable poin-in-time-recovery (PITR) you should consider log archiving. It should also save you heaps of load on cpu, disk, network and postgresql server.
-Amos
On 11/17/09, Benjamin Smith lists@benjamindsmith.com wrote:
I'm having a server report a high load average when backing up Postgres database files to an external USB drive. This is driving my loadbalancers all out of kilter and causing a large volume of network monitor alerts.
I have a 1TB USB drive plugged into a USB2 port that I use to back up the production drives (which are SCSI). It's working fine, but while doing backups (hourly) the load average on the server shoots up from the normal 0.5 - 1.5 or so up to a high between 10 and 30. Strangely, even though the "load is high" the server is completely responsive, even the USB drives being accessed are!
Backup script is really simple, run via cron, pretty much just:
#! /bin/sh hour=`date +%k`; pg_dump <options> mydatabase > /media/backups/mydatabase.$hour.pgsql;
where /media/backups is the mount point for the USB drive.
Using top to diagnose, nothing seems to be particularly high! IoWait seems reasonable (10-30%) and CPUs are 0.5%, Idle is 70-90%. Even accessing the USB partition while the load is "high" is responsive!
I'm guessing that something changed in how load average is counted?
Server Stats: Late model 8-way Xeon, SuperMicro brand. CentOS 4.x / 64 (all updates applied, booted after last kernel update) Kernel 2.6.9-89.0.16.ELsmp 4 GB ECC RAM 300 GB SCSI HDD. Standard Apache/PHP, Postgres 8.4.
Any idea how to revert to the old load average tracking behavior short of using a stale and potentially insecure kernel?
-- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-
Amos Shapira -
Benjamin Smith -
Todd Denniston