scripts: add some metrics for tidb

[jackysp]

Signed-off-by: Shuaipeng Yu [email protected]
This script shows what it looks like http://172.16.5.34:34125/d/000000011/test-cluster-tidb?orgId=1&from=now-1h&to=now&refresh=30s

[tiancaiamao]

Do we have slow_query related metrics? @jackysp

[jackysp]

Yes, already have one.

[lysu]

how about add

• process_open_fds identify socket fd problem(mainly to tikv side connection increase)
• go_memstats_gc_cpu_fraction percent between 0 to 1
• go_memstats_heap_inuse maybe need change to go_memstats_heap_alloc_bytes? https://golang.org/pkg/runtime/#MemStats(maybe we can copy comments to metric's description)
• sum(tidb_tikvclient_pending_batch_requests) by (instance) => sum(tidb_tikvclient_pending_batch_requests) by (store)
• maybe we need add tidb_tikvclient_lock_cleanup_task_total and tidb_tikvclient_range_task_stats with its labels?

[lysu]

rest LGTM

[jackysp]

how about add

• process_open_fds identify socket fd problem(mainly to tikv side connection increase)
  good idea
• go_memstats_gc_cpu_fraction percent between 0 to 1
  ok
• go_memstats_heap_inuse maybe need change to go_memstats_heap_alloc_bytes? https://golang.org/pkg/runtime/#MemStats(maybe we can copy comments to metric's description)
  ok, but I firstly do not want to change the existed formula.
• sum(tidb_tikvclient_pending_batch_requests) by (instance) => sum(tidb_tikvclient_pending_batch_requests) by (store)
  ok, but I'm afraid the num of tikv will be so many.
• maybe we need add tidb_tikvclient_lock_cleanup_task_total and tidb_tikvclient_range_task_stats with its labels?
  They're already there.

[jackysp]

PTAL @lysu

[lysu]

LGTM

[zz-jason]

LGTM