Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.Original topic: tikv detail grafana中flow control是什么意思？
|
username: lindoubled
Version:
v5.3.0
What does the above image mean? It’s not documented.
Question:
When Throttle duration appears, TiDB experiences many slow queries. How can this be optimized?
This should refer to TiKV’s server is busy flow control. You can further confirm it in the TiKV-Details → Errors → Server is busy monitoring. For more details, you can refer to the following document:
By checking the monitoring: Grafana → TiKV → errors to confirm the specific busy reason.
server is busyis TiKV’s own flow control mechanism. TiKV uses this method to informtidb/ti-clientthat the current TiKV is under too much pressure and to try again later.
That means the flow control was not triggered, but this page is mainly used to monitor flow control, including write stall, compaction pile-up, L0 SST pile-up, etc. You can specifically check which command’s delay has increased on the Throttle duration.
The commit log includes network and disk writing, check these two aspects.
It doesn’t feel right; there’s also a commit log duration here.
Why are these metrics so messy?
Flow control is the monitoring of scheduler flow control after version 5.2. Check if storage.flow-control is enabled. There is indeed relatively little specific description in the official documentation.
Check the network status under black exporter/node exporter monitoring.
Each instance is almost like this, with a 10 Gigabit network card.
Blackexporter has network latency, but it is also at a 30-second granularity.
This topic was automatically closed 1 minute after the last reply. No new replies are allowed.
|
username: 
|
username: 
|
username: