Обсуждение: Gigantic load average spikes
Postgres 10.3 On freebsd 10.3 is almost idle, disk i/o about 5-10%, number running processes about 1-3, cpu about 90% idle, then we run a i/o heavy job like "refresh materialized view", cpu & disk i/o still not maxed out, all of a sudden the number of running processes increases to about 250-300 for a second, which increases load averages that we periodically poll to see if everything's alright. Why is that? Disk i/o is far from being maxed out, the actual number of running processes is really small.
On 3/31/19 8:23 AM, rihad wrote: > Postgres 10.3 On freebsd 10.3 is almost idle, disk i/o about 5-10%, > number running processes about 1-3, cpu about 90% idle, then we run a > i/o heavy job like "refresh materialized view", cpu & disk i/o still not > maxed out, all of a sudden the number of running processes increases to > about 250-300 for a second, which increases load averages that we Do you know what the processes are? Are there any messages in the Postgres log from that time period that might shed light? > periodically poll to see if everything's alright. Why is that? Disk i/o > is far from being maxed out, the actual number of running processes is > really small. You are referring to the between spikes process count? > > > > -- Adrian Klaver adrian.klaver@aklaver.com
On 2019-03-31 19:23:11 +0400, rihad wrote:
> Postgres 10.3 On freebsd 10.3 is almost idle, disk i/o about 5-10%, number
> running processes about 1-3, cpu about 90% idle, then we run a i/o heavy job
> like "refresh materialized view", cpu & disk i/o still not maxed out, all of
> a sudden the number of running processes increases to about 250-300 for a
What exactly do you mean by "running processes"? I don't think I've ever
seen a Unix with only 1 to 3 running processes in total, so you are
probably referring to processes in a certain state. Runnable (R)?
Uninterruptible sleep (D)? Both? Something else?
hp
--
_ | Peter J. Holzer | we build much bigger, better disasters now
|_|_) | | because we have much more sophisticated
| | | hjp@hjp.at | management tools.
__/ | http://www.hjp.at/ | -- Ross Anderson <https://www.edge.org/>
Вложения
> What exactly do you mean by "running processes"? I don't think I've ever > seen a Unix with only 1 to 3 running processes in total, so you are > probably referring to processes in a certain state. Runnable (R)? > Uninterruptible sleep (D)? Both? Something else? Just that, 250-300 running processes in top shown for a second. 250-300 is the number of postgres worker processes used, but normally only 1-3 of them are running according to top. At times of load FreeBSD (?) schedules all of them to run. This doesn't really put the machine on its knees, it just impacts load avg.
On Mon, 1 Apr 2019 at 18:08, rihad <rihad@mail.ru> wrote: > > > What exactly do you mean by "running processes"? I don't think I've ever > > seen a Unix with only 1 to 3 running processes in total, so you are > > probably referring to processes in a certain state. Runnable (R)? > > Uninterruptible sleep (D)? Both? Something else? > > Just that, 250-300 running processes in top shown for a second. 250-300 > is the number of postgres worker processes used, but normally only 1-3 > of them are running according to top. At times of load FreeBSD (?) > schedules all of them to run. This doesn't really put the machine on its > knees, it just impacts load avg. Perhaps a bunch of processes waiting on the access exclusive lock on the materialized view being released? log_lock_waits might help you if the MV takes more than a second to refresh, otherwise, you might need to have a look at ungranted locks in pg_locks and see if the number of locks spikes during the refresh. -- David Rowley http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training & Services
On 3/31/19 10:08 PM, rihad wrote: >> What exactly do you mean by "running processes"? I don't think I've ever >> seen a Unix with only 1 to 3 running processes in total, so you are >> probably referring to processes in a certain state. Runnable (R)? >> Uninterruptible sleep (D)? Both? Something else? > > Just that, 250-300 running processes in top shown for a second. 250-300 > is the number of postgres worker processes used, but normally only 1-3 > of them are running according to top. At times of load FreeBSD (?) So what are the process titles? > schedules all of them to run. This doesn't really put the machine on its > knees, it just impacts load avg. > > > > -- Adrian Klaver adrian.klaver@aklaver.com
On 04/01/2019 06:17 PM, Adrian Klaver wrote: > On 3/31/19 10:08 PM, rihad wrote: >>> What exactly do you mean by "running processes"? I don't think I've >>> ever >>> seen a Unix with only 1 to 3 running processes in total, so you are >>> probably referring to processes in a certain state. Runnable (R)? >>> Uninterruptible sleep (D)? Both? Something else? >> >> Just that, 250-300 running processes in top shown for a second. >> 250-300 is the number of postgres worker processes used, but normally >> only 1-3 of them are running according to top. At times of load >> FreeBSD (?) > > So what are the process titles? postgres
On 4/1/19 8:06 AM, rihad wrote: > On 04/01/2019 06:17 PM, Adrian Klaver wrote: >> On 3/31/19 10:08 PM, rihad wrote: >>>> What exactly do you mean by "running processes"? I don't think I've >>>> ever >>>> seen a Unix with only 1 to 3 running processes in total, so you are >>>> probably referring to processes in a certain state. Runnable (R)? >>>> Uninterruptible sleep (D)? Both? Something else? >>> >>> Just that, 250-300 running processes in top shown for a second. >>> 250-300 is the number of postgres worker processes used, but normally >>> only 1-3 of them are running according to top. At times of load >>> FreeBSD (?) >> >> So what are the process titles? > > postgres Have you tried hitting the 'c' key in top to get a fuller description? > > -- Adrian Klaver adrian.klaver@aklaver.com
On Sun, Mar 31, 2019 at 10:49 PM David Rowley <david.rowley@2ndquadrant.com> wrote:
Perhaps a bunch of processes waiting on the access exclusive lock on
the materialized view being released?
log_lock_waits might help you if the MV takes more than a second to
refresh, otherwise, you might need to have a look at ungranted locks
in pg_locks and see if the number of locks spikes during the refresh.
I think David's got the right idea here. Like he said, investigate pg_locks, if it is the refresh materialized view, you can avoid the problem by doing 'REFRESH MATERIALIZED VIEW CONCURRENTLY'. You will need at least one unique index on the table.
--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
On 04/01/2019 08:30 PM, Michel Pelletier wrote:
On Sun, Mar 31, 2019 at 10:49 PM David Rowley <david.rowley@2ndquadrant.com> wrote:
Perhaps a bunch of processes waiting on the access exclusive lock on
the materialized view being released?
log_lock_waits might help you if the MV takes more than a second to
refresh, otherwise, you might need to have a look at ungranted locks
in pg_locks and see if the number of locks spikes during the refresh.I think David's got the right idea here. Like he said, investigate pg_locks, if it is the refresh materialized view, you can avoid the problem by doing 'REFRESH MATERIALIZED VIEW CONCURRENTLY'. You will need at least one unique index on the table.
It is actually refreshed concurrently.
--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
On Mon, Apr 1, 2019 at 10:35 AM rihad <rihad@mail.ru> wrote:
On 04/01/2019 08:30 PM, Michel Pelletier wrote:On Sun, Mar 31, 2019 at 10:49 PM David Rowley <david.rowley@2ndquadrant.com> wrote:
Perhaps a bunch of processes waiting on the access exclusive lock on
the materialized view being released?
log_lock_waits might help you if the MV takes more than a second to
refresh, otherwise, you might need to have a look at ungranted locks
in pg_locks and see if the number of locks spikes during the refresh.I think David's got the right idea here. Like he said, investigate pg_locks, if it is the refresh materialized view, you can avoid the problem by doing 'REFRESH MATERIALIZED VIEW CONCURRENTLY'. You will need at least one unique index on the table.
It is actually refreshed concurrently.
--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services