*: Add dbSizeInUse to StatusResposne #9256
Conversation
etcdserver/etcdserverpb/rpc.proto
Outdated
| @@ -911,6 +911,8 @@ message StatusResponse { | |||
| uint64 raftAppliedIndex = 7; | |||
| // errors contains alarm/health information and status. | |||
| repeated string errors = 8; | |||
| // dbSize is the size of the backend database logically in use, in bytes, of the responding member. | |||
There was a problem hiding this comment.
i think we should add this as a metrics too. not just putting this into the status response.
There was a problem hiding this comment.
Fixed.
Also added a metrics.
| @@ -300,10 +300,13 @@ func (s *store) Restore(b backend.Backend) error { | |||
| } | |||
|
|
|||
| func (s *store) restore() error { | |||
| reportDbTotalSizeInBytesMu.Lock() | |||
| b := s.b | |||
There was a problem hiding this comment.
I couldn't find a particular reason to procted b := s.b with the mutex. Let me know if I overlooked something.
yudai
force-pushed
the
actual_db_size
branch
2 times, most recently
from
eb5c4b5 to
fef9309
Compare
internal/mvcc/backend/backend.go
Outdated
| @@ -54,6 +54,7 @@ type Backend interface { | |||
| Hash(ignores map[IgnoreKey]struct{}) (uint32, error) | |||
| // Size returns the current size of the backend. | |||
| Size() int64 | |||
| SizeInUse() int64 | |||
There was a problem hiding this comment.
comment on this interface method?
|
look good to me overall! we need to update https://github.com/coreos/etcd/blob/master/Documentation/op-guide/maintenance.md#defragmentation to mention the in-use-db-size to identify if a defrag is needed. |
|
Also update changelog https://github.com/coreos/etcd/blob/master/CHANGELOG-3.4.md? e.g. https://github.com/coreos/etcd/blob/master/CHANGELOG-3.3.md#addedmetrics Thanks! |
yudai
force-pushed
the
actual_db_size
branch
2 times, most recently
from
4a8ccda to
03054b8
Compare
|
Thanks for your comments. Updated the code (just let me know or overwrite in case it has English errors). I'm wondering if we want to show the value by |
Let us do it in another PR? |
Got it. Somehow a test failed due to timeout. Probably #9086. |
CHANGELOG-3.4.md
Outdated
| @@ -51,6 +51,8 @@ | |||
| - Add [`raftAppliedIndex` field to `etcdserverpb.StatusResponse`](https://github.com/coreos/etcd/pull/9176) for current Raft applied index. | |||
| - Add [`errors` field to `etcdserverpb.StatusResponse`](https://github.com/coreos/etcd/pull/9206) for server-side error. | |||
| - e.g. `"etcdserver: no leader", "NOSPACE", "CORRUPT"` | |||
| - Add [`dbSizeInUse` field to `etcdserverpb.StatusResponse`](https://github.com/coreos/etcd/pull/9256) for actual DB size after compaction. | |||
| - Also exposed as metrics `etcd_debugging_mvcc_db_total_size_in_use_in_bytes` | |||
Existing dbSize shows physically allocated DB size and the backend (boltdb) won't shrink it after a compaction until a user runs the defrag command. The new dbSizeInUse shows the DB size that excludes free pages created by compactions so that users can see the actual DB usage. dbSize >= dbSizeInUse is always true. Note that dbSizeInUse shows a page-based size and not byte level usage.
Existing dbSize shows physically allocated DB size and the backend
(boltdb) won't shrink it after a compaction until a user runs the defrag command.
The new dbSizeInUse shows the DB size that excludes free pages created
by compactions so that users can see the actual DB usage. dbSize >=
dbSizeInUse is always true.
Note that dbSizeInUse shows a page-based size and not byte level usage.
For #9222.