Hi All,
Sorry for the long post, but we're a bit stuck!
We have a situation where we get a slowdown of the database. This seems to happen randomly during the month/year, but always at the same time range during the day.
Our customer has a workgroup license (10.1b) and database size is around 20gb. We've just moved them to a brand new server which we hoped would solve the issue - it's RHEL 6, solid state disks, Intel Xeon 24 cores, 16Gb ram).
We now have a process that detects the slowdown and notifies us, but still can't pinpoint what it is.
Our detection process simply does a for each of the customer table to get 2000 records and times the start and finish, and it does this every 5 seconds:
using this, a database at rest normally returns a difference of 9-13ms, but during the slowdown, it can take anywhere between 20000-30000ms, which is a huge difference! the script records the last 3 time differences, and if all 3 are over 500ms, we get notified (to avoid spikes).
Also, during the night, we have a statistical batch process that hits the database very hard for about 3-4 hours, and that doesn't get picked up, so whatever this is, it's very severe.
We contacted Progress, but all they could offer was to increase the -B and change the -spin value (but that's only for Enterprise).
Looking at promon during the slowdown, the buffers are at 85%-95%, and every other database responds as expected.
The BI stays more or less the same size, and disk monitoring and CPU utils don't report anything out of the ordinary, and there are no linux cron jobs running during the time frame. Apart from the database, everything on the system responds nicely.
Our only conclusion is that it's something to do with the database engine itself, but other than that we are lost.
Any idea what else I could check or change to help find the issue would be appreciated.
Sorry for the long post, but we're a bit stuck!
We have a situation where we get a slowdown of the database. This seems to happen randomly during the month/year, but always at the same time range during the day.
Our customer has a workgroup license (10.1b) and database size is around 20gb. We've just moved them to a brand new server which we hoped would solve the issue - it's RHEL 6, solid state disks, Intel Xeon 24 cores, 16Gb ram).
We now have a process that detects the slowdown and notifies us, but still can't pinpoint what it is.
Our detection process simply does a for each of the customer table to get 2000 records and times the start and finish, and it does this every 5 seconds:
Code:
v_start = etime.
for each customer
no-lock
x = 1 to 2000
:
end.
v_end = etime.
Also, during the night, we have a statistical batch process that hits the database very hard for about 3-4 hours, and that doesn't get picked up, so whatever this is, it's very severe.
We contacted Progress, but all they could offer was to increase the -B and change the -spin value (but that's only for Enterprise).
Looking at promon during the slowdown, the buffers are at 85%-95%, and every other database responds as expected.
The BI stays more or less the same size, and disk monitoring and CPU utils don't report anything out of the ordinary, and there are no linux cron jobs running during the time frame. Apart from the database, everything on the system responds nicely.
Our only conclusion is that it's something to do with the database engine itself, but other than that we are lost.
Any idea what else I could check or change to help find the issue would be appreciated.