Waits useless on MySQL?
From: kyle Hailey <kylelf_at_gmail.com>
Date: Mon, 23 Jan 2023 15:18:37 -0800
Message-ID: <CADsdiQhFT5ZB4RzvmmjqJq7hNhhL6Wye18qTf9FGkK5U+GLJjA_at_mail.gmail.com>
Anyone tried to use wait events on MySQL? Looking across a large fleet of MySQL, 70% of the wait time is on wait/io/table/sql/handler.
The event wait/io/table/sql/handler can be CPU, IO and lock! Seems like waits are pretty useless.
We captured them in RDS and are currently doing so at Datadog, but thinking might even be worth skipping their collections and just reporting MySQL state instead of waits.
Seems incredible that row level locks are not instrumented as their own waits. (Looks like we can deduce lock in rather unsatisfying way TBD)
Date: Mon, 23 Jan 2023 15:18:37 -0800
Message-ID: <CADsdiQhFT5ZB4RzvmmjqJq7hNhhL6Wye18qTf9FGkK5U+GLJjA_at_mail.gmail.com>
Anyone tried to use wait events on MySQL? Looking across a large fleet of MySQL, 70% of the wait time is on wait/io/table/sql/handler.
The event wait/io/table/sql/handler can be CPU, IO and lock! Seems like waits are pretty useless.
We captured them in RDS and are currently doing so at Datadog, but thinking might even be worth skipping their collections and just reporting MySQL state instead of waits.
Seems incredible that row level locks are not instrumented as their own waits. (Looks like we can deduce lock in rather unsatisfying way TBD)
[image: image.png]
-- http://www.freelists.org/webpage/oracle-lReceived on Tue Jan 24 2023 - 00:18:37 CET
- image/png attachment: image.png