[google-cloud-sql-discuss] Re: Can I drop the failover CloudSQL instance and re-create it?

Hi,

If I see the the logs for my database on the "Logs Viewer", there are following kind of messages that seem to be appearing almost 4 minutes apart each time:

InnoDB: page_cleaner: 1000ms intended loop took 13945ms. The settings might not be optimal.

Could the two things be interrelated (process of flushing the buffers to disk and drop in queries performance at the same time)?

Can I access the CloudSql logs in real-time on console somehow? Not sure whether on Stackdriver Logs Viewer they appear immediately or after some lag. It'll be interesting to see them on console to check whether the timing of the two things coincides.

Cheers,

Roshan

On Wednesday, 2 May 2018 12:58:50 UTC+5:30, Roshan Dawrani wrote:

Hi Fady,

Thanks a lot for replying.

I checked and the replica lag for my instance is set to 0 (zero). I suppose that means that failover is almost being kept instantaneously in sync with the main instance?

So, you are saying that keeping failover instance in-sync couldn't be the reason for the drop I am (almost religiously) seeing every 4-5 minutes in the performance of the queries (both read and write)?

1) Is there any ideas you have on how I may dig up further to figure what might be happening on the db instance at such regular intervals? Could there be some MySQL level compaction, etc triggering at these 4-5 minutes interval? How do I start digging up such an issue?

2) Since the pausing of failover is not possible, are there any huge downsides to deleting and re-creating the replica (apart from the fact that for the duration of experimentation, the instance will not have a failover backup)?

     * When the failover is re-introduced, how will it be re-constructed? Will it be gradual, or are we looking at a db-performance risk because it'll be working too hard to get the failover up again? Our db is not on the small side, so I am a little worried about the effort it'll need to put in to get a new failover up-to-date.

Cheers,

Roshan

On Tuesday, 1 May 2018 02:18:00 UTC+5:30, Fady (Google Cloud Platform) wrote:

Hello Roshan,

According to this document " To minimize performance impact on the master, while ensuring that changes are never lost, the replica logs the update events, and then performs the updates in order."  Therefore, I highly doubt that there is specific task that runs every 4- 5 minutes to keep both instances synced, but rather in seconds per order of each operation . You may verify that by checking the replication lag metrics (seconds behind master).  

As for stopping/pausing the failover replica, it is not currently possible. As you mentioned, you would have to delete it, and this would render your master not to be configured with high availability. If you choose to recreate the replica, you may check this guide for quick configurations. I hope this helps

--
You received this message because you are subscribed to the Google Groups "Google Cloud SQL discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to google-cloud-sql-discuss+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/google-cloud-sql-discuss/5ce1c653-64e7-4d8e-8782-2cc3f3fdbc37%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.