From: | Joshua Drake <jd(at)commandprompt(dot)com> |
---|---|
To: | Don Seiler <don(at)seiler(dot)us> |
Cc: | pgsql-performance(at)lists(dot)postgresql(dot)org |
Subject: | Re: High COMMIT times |
Date: | 2021-01-06 18:53:27 |
Message-ID: | CAJvJg-QgXHG9yXbtyLMFmUU5rZ2wPnbCxFQb7QSH7o41nFBV8A@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
>
> Looking at the Azure portal metric, we are nowhere close to the advertised
> maximum IOPS or MB/s throughput (under half of the maximum IOPS and under a
> quarter of the MB/s maximum). So there must be some other bottleneck in
> play. The IOPS limit on this VM size is even higher so that shouldn't be it.
>
> If I were to size a separate volume for just WAL, I would think 64GB but
> obviously the Azure storage IOPS are much less. On this particular DB host,
> we're currently on a 2.0T P40 disk that is supposed to give 7500 IOPS and
> 250MB/s [1] (but again, Azure's own usage graphs show us nowhere near those
> limits). A smaller volume like 64GB would be provisioned at 240 IOPS in
> this example. Doesn't seem like a lot. Really until you get a terabyte it
> seems like a real drop-off as far as Azure storage goes.
>
>
Based on those metrics, I would start looking at other things. For example,
I once (it was years ago) experienced commit delays because the kernel
cache on Linux was getting over run. Do you have any metrics on the system
as a whole? Perhaps sar running every few minutes will help you identify a
correlation?
JD
> I'd be interested to hear what others might have configured on their
> write-heavy cloud databases.
>
> [1] https://azure.microsoft.com/en-us/pricing/details/managed-disks/
>
> Don.
>
> --
> Don Seiler
> www.seiler.us
>
From | Date | Subject | |
---|---|---|---|
Next Message | Kenneth Marshall | 2021-01-06 19:15:25 | Re: High COMMIT times |
Previous Message | Don Seiler | 2021-01-06 18:06:27 | Re: High COMMIT times |