Thread: Vacuum duration + hint bits?
Hi, I have a question about a db-wide vacuum that I am running that is taking a much longer time than normal. We switched over to our warm standby server today -- which is virtually identical to the source db server -- and I initiated a "vacuum analyze verbose". Normally this process wouldn't take more than 6 hours, but so far we are well over 9 hours. I seem to recall reading in one of these pg lists that a db-wide vacuum of a new (?) database would go about setting the hint bits and cause a lot more disk IO than is normal. Is that a possible cause.
Postgresql 8.2.13, RHEL5
Sorry, I realize I'm short on details, but I'm just running out the door and the thought about hint bits struck me.
Cheers!
Postgresql 8.2.13, RHEL5
Sorry, I realize I'm short on details, but I'm just running out the door and the thought about hint bits struck me.
Cheers!
bricklen <bricklen@gmail.com> writes: > Hi, I have a question about a db-wide vacuum that I am running that is > taking a much longer time than normal. We switched over to our warm standby > server today -- which is virtually identical to the source db server -- and > I initiated a "vacuum analyze verbose". Normally this process wouldn't take > more than 6 hours, but so far we are well over 9 hours. I seem to recall > reading in one of these pg lists that a db-wide vacuum of a new (?) database > would go about setting the hint bits and cause a lot more disk IO than is > normal. Is that a possible cause. Yeah, it seems possible. You could look at vmstat to see if there's tons of write activity ... regards, tom lane
Yeah, there's a lot. Way more than I am accustomed to seeing from the same command on the previous server.
On Thu, Aug 27, 2009 at 4:05 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
bricklen <bricklen@gmail.com> writes:Yeah, it seems possible. You could look at vmstat to see if there's
> Hi, I have a question about a db-wide vacuum that I am running that is
> taking a much longer time than normal. We switched over to our warm standby
> server today -- which is virtually identical to the source db server -- and
> I initiated a "vacuum analyze verbose". Normally this process wouldn't take
> more than 6 hours, but so far we are well over 9 hours. I seem to recall
> reading in one of these pg lists that a db-wide vacuum of a new (?) database
> would go about setting the hint bits and cause a lot more disk IO than is
> normal. Is that a possible cause.
tons of write activity ...
regards, tom lane