You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
suggest to add an parameter check-interval to control the interval of check load and slave lag. #166
suggest to add an parameter check-interval to control the interval of check load and slave lag.
I read the code and find the interval of throttler is set to 1 second, in some situation, the load and lag must check more frequently.
This would be an easy fix. I'm curious about needing to poll replicas more frequently than once per 1sec, given that replication lag resolution in MySQL is 1sec. Can you please elaborate?
OK, thank you for your suggest.
We use mysql in some application which is very sensitive about slave lag, I have suggest the RD use master in this situation, but sometimes they forget, and they want the as small lag as possible, especially in the tools.
I don't understand very well.but seconds_behind_master is only in second, so I think 1s is enough. And check the pt-osc check-interval default value is 5s, we set to 1s when we use.
I'm curious about needing to poll replicas more frequently than once per 1sec, given that replication lag resolution in MySQL is 1sec
I stand corrected, and we are able (on 5.6) to get subsecond granularity. We are good to go on polling for replication lag smaller than 1s and in shorter interval. See #203 (comment) for more on this.
We're unlikely to sample max-load more frequently.
Activity
shlomi-noach commentedon Aug 18, 2016
pro tip #1: use three backtics for multi line comments (see how I edited your comment)
pro tip #2: you can link to a specific line in code: https://github.com/github/gh-ost/blob/master/go/logic/migrator.go#L191
This would be an easy fix. I'm curious about needing to poll replicas more frequently than once per 1sec, given that replication lag resolution in MySQL is 1sec. Can you please elaborate?
rj03hou commentedon Aug 19, 2016
OK, thank you for your suggest.
We use mysql in some application which is very sensitive about slave lag, I have suggest the RD use master in this situation, but sometimes they forget, and they want the as small lag as possible, especially in the tools.
shlomi-noach commentedon Aug 19, 2016
Cool, but what's the deal of more than once per second in light of the resolution?
rj03hou commentedon Aug 19, 2016
I don't understand very well.but seconds_behind_master is only in second, so I think 1s is enough. And check the pt-osc check-interval default value is 5s, we set to 1s when we use.
shlomi-noach commentedon Aug 19, 2016
I'm closing this for now. If there's a particular case for reducing polling interval, either reopen this or open a new one.
shlomi-noach commentedon Aug 29, 2016
I stand corrected, and we are able (on
5.6) to get subsecond granularity. We are good to go on polling for replication lag smaller than1sand in shorter interval. See #203 (comment) for more on this.We're unlikely to sample
max-loadmore frequently.shlomi-noach commentedon Aug 31, 2016
Closed as per #203 (comment)