*: document compactor changes

Signed-off-by: Gyuho Lee <gyuhox@gmail.com>

CHANGELOG-3.3.md

@@ -9,15 +9,15 @@ See [code changes](https://github.com/coreos/etcd/compare/v3.3.2...v3.3.3) and [
 - Adjust [election timeout on server restart](https://github.com/coreos/etcd/pull/9415) to reduce [disruptive rejoining servers](https://github.com/coreos/etcd/issues/9333).
   - Previously, etcd fast-forwards election ticks on server start, with only one tick left for leader election. This is to speed up start phase, without having to wait until all election ticks elapse. Advancing election ticks is useful for cross datacenter deployments with larger election timeouts. However, it was affecting cluster availability if the last tick elapses before leader contacts the restarted node.
   - Now, when etcd restarts, it adjusts election ticks with more than one tick left, thus more time for leader to prevent disruptive restart.
-
-### Fixed: v3
-
-- TODO: Fix [compaction interval calculation](TODO).
-  - Previously, `--auto-compaction-mode=periodic --auto-compaction-retention=10h` automatically `Compact` on latest revision at first 10-hour and every 1-hour, whether it succeeds or not.
-  - Now, it correctly reset its interval when `Compact` operation succeeds.
-  - The failed `Compact` operation will still be retried in 1/10 of interval.
-  - e.g. `10h` compact interval will be retried in 1-hour on failure.
-  - Document...
+- Adjust [periodic compaction retention window](https://github.com/coreos/etcd/pull/9474).
+  - e.g. `--auto-compaction-mode=revision --auto-compaction-retention=1000` automatically `Compact` on `"latest revision" - 1000` every 5-minute (when latest revision is 30000, compact on revision 29000).
+  - e.g. Previously, `--auto-compaction-mode=periodic --auto-compaction-retention=24h` automatically `Compact` with 24-hour retention windown for every 2.4-hour. Now, `Compact` happens for every 1-hour.
+  - e.g. Previously, `--auto-compaction-mode=periodic --auto-compaction-retention=30m` automatically `Compact` with 30-minute retention windown for every 3-minute. Now, `Compact` happens for every 30-minute.
+  - Periodic compactor keeps recording latest revisions for every compaction period when given period is less than 1-hour, or for every 1-hour when given compaction period is greater than 1-hour (e.g. 1-hour when `--auto-compaction-mode=periodic --auto-compaction-retention=24h`).
+  - For every compaction period or 1-hour, compactor uses the last revision that was fetched before compaction period, to discard historical data.
+  - The retention window of compaction period moves for every given compaction period or hour.
+  - For instance, when hourly writes are 100 and `--auto-compaction-mode=periodic --auto-compaction-retention=24h`, `v3.2.x`, `v3.3.0`, `v3.3.1`, and `v3.3.2` compact revision 2400, 2640, and 2880 for every 2.4-hour, while `v3.3.3` *or later* compacts revision 2400, 2500, 2600 for every 1-hour.
+  - Futhermore, when `--auto-compaction-mode=periodic --auto-compaction-retention=30m` and writes per minute are about 1000, `v3.3.0`, `v3.3.1`, and `v3.3.2` compact revision 30000, 33000, and 36000, for every 3-minute, while `v3.3.3` *or later* compacts revision 30000, 60000, and 90000.
 
 ### Metrics, Monitoring
 
@@ -179,12 +179,13 @@ See [security doc](https://github.com/coreos/etcd/blob/master/Documentation/op-g
   - Useful for [bypassing critical APIs when monitoring etcd](https://github.com/coreos/etcd/issues/8060).
 - Add [`--auto-compaction-mode`](https://github.com/coreos/etcd/pull/8123) flag to [support revision-based compaction](https://github.com/coreos/etcd/issues/8098).
 - Change `--auto-compaction-retention` flag to [accept string values](https://github.com/coreos/etcd/pull/8563) with [finer granularity](https://github.com/coreos/etcd/issues/8503).
-  - e.g. `--auto-compaction-mode=periodic --auto-compaction-retention=30m` automatically `Compact` on latest revision every 30-minute.
   - e.g. `--auto-compaction-mode=revision --auto-compaction-retention=1000` automatically `Compact` on `"latest revision" - 1000` every 5-minute (when latest revision is 30000, compact on revision 29000).
+  - e.g. `--auto-compaction-mode=periodic --auto-compaction-retention=24h` automatically `Compact` with 24-hour retention windown for every 2.4-hour.
+  - e.g. `--auto-compaction-mode=periodic --auto-compaction-retention=30m` automatically `Compact` with 30-minute retention windown for every 3-minute.
   - Periodic compactor continues to record latest revisions for every 1/10 of given compaction period (e.g. 1-hour when `--auto-compaction-mode=periodic --auto-compaction-retention=10h`).
   - For every 1/10 of given compaction period, compactor uses the last revision that was fetched before compaction period, to discard historical data.
   - The retention window of compaction period moves for every 1/10 of given compaction period.
-  - For instance, when hourly writes are about 100 and `--auto-compaction-retention=10`, v3.1 compacts revision 1000, 2000, and 3000 for every 10-hour, while v3.2 and v3.3 compacts revision 1000, 1100, and 1200 for every 1-hour. Futhermore, when writes per minute are about 1000, v3.3 with `--auto-compaction-mode=periodic --auto-compaction-retention=30m` can compact revision 30000, 33000, and 36000, for every 3-minute with more finer granularity.
+  - For instance, when hourly writes are 100 and `--auto-compaction-retention=10`, v3.1 compacts revision 1000, 2000, and 3000 for every 10-hour, while v3.2.x, v3.3.0, v3.3.1, and v3.3.2 compact revision 1000, 1100, and 1200 for every 1-hour. Futhermore, when writes per minute are 1000, v3.3.0, v3.3.1, and v3.3.2 with `--auto-compaction-mode=periodic --auto-compaction-retention=30m` compact revision 30000, 33000, and 36000, for every 3-minute with more finer granularity.
   - Whether compaction succeeds or not, this process repeats for every 1/10 of given compaction period. If compaction succeeds, it just removes compacted revision from historical revision records.
 - Add [`--grpc-keepalive-min-time`, `--grpc-keepalive-interval`, `--grpc-keepalive-timeout`](https://github.com/coreos/etcd/pull/8535) flags to configure server-side keepalive policies.
 - Serve [`/health` endpoint as unhealthy](https://github.com/coreos/etcd/pull/8272) when [alarm (e.g. `NOSPACE`) is raised or there's no leader](https://github.com/coreos/etcd/issues/8207).

CHANGELOG-3.4.md

@@ -18,6 +18,15 @@ See [code changes](https://github.com/coreos/etcd/compare/v3.3.0...v3.4.0) and [
   - Now, when etcd restarts, it adjusts election ticks with more than one tick left, thus more time for leader to prevent disruptive restart.
 - Add [Raft Pre-Vote feature](https://github.com/coreos/etcd/pull/9352) to reduce [disruptive rejoining servers](https://github.com/coreos/etcd/issues/9333).
   - For instance, a flaky(or rejoining) member may drop in and out, and start campaign. This member will end up with a higher term, and ignore all incoming messages with lower term. In this case, a new leader eventually need to get elected, thus disruptive to cluster availability. Raft implements Pre-Vote phase to prevent this kind of disruptions. If enabled, Raft runs an additional phase of election to check if pre-candidate can get enough votes to win an election.
+- Adjust [periodic compaction retention window](https://github.com/coreos/etcd/pull/9474).
+  - e.g. `--auto-compaction-mode=revision --auto-compaction-retention=1000` automatically `Compact` on `"latest revision" - 1000` every 5-minute (when latest revision is 30000, compact on revision 29000).
+  - e.g. Previously, `--auto-compaction-mode=periodic --auto-compaction-retention=24h` automatically `Compact` with 24-hour retention windown for every 2.4-hour. Now, `Compact` happens for every 1-hour.
+  - e.g. Previously, `--auto-compaction-mode=periodic --auto-compaction-retention=30m` automatically `Compact` with 30-minute retention windown for every 3-minute. Now, `Compact` happens for every 30-minute.
+  - Periodic compactor keeps recording latest revisions for every compaction period when given period is less than 1-hour, or for every 1-hour when given compaction period is greater than 1-hour (e.g. 1-hour when `--auto-compaction-mode=periodic --auto-compaction-retention=24h`).
+  - For every compaction period or 1-hour, compactor uses the last revision that was fetched before compaction period, to discard historical data.
+  - The retention window of compaction period moves for every given compaction period or hour.
+  - For instance, when hourly writes are 100 and `--auto-compaction-mode=periodic --auto-compaction-retention=24h`, `v3.2.x`, `v3.3.0`, `v3.3.1`, and `v3.3.2` compact revision 2400, 2640, and 2880 for every 2.4-hour, while `v3.3.3` *or later* compacts revision 2400, 2500, 2600 for every 1-hour.
+  - Futhermore, when `--auto-compaction-mode=periodic --auto-compaction-retention=30m` and writes per minute are about 1000, `v3.3.0`, `v3.3.1`, and `v3.3.2` compact revision 30000, 33000, and 36000, for every 3-minute, while `v3.3.3` *or later* compacts revision 30000, 60000, and 90000.
 - Make [Lease `Lookup` non-blocking with concurrent `Grant`/`Revoke`](https://github.com/coreos/etcd/pull/9229).
 
 ### Breaking Changes
@@ -152,12 +161,6 @@ See [security doc](https://github.com/coreos/etcd/blob/master/Documentation/op-g
 - Fix [revision-based compaction retention parsing](https://github.com/coreos/etcd/pull/9339).
   - Previously, `etcd --auto-compaction-mode revision --auto-compaction-retention 1` was [translated to revision retention 3600000000000](https://github.com/coreos/etcd/issues/9337).
   - Now, `etcd --auto-compaction-mode revision --auto-compaction-retention 1` is correctly parsed as revision retention 1.
-- TODO: Fix [compaction interval calculation](TODO).
-  - Previously, `--auto-compaction-mode=periodic --auto-compaction-retention=10h` automatically `Compact` on latest revision at first 10-hour and every 1-hour, whether it succeeds or not.
-  - Now, it correctly reset its interval when `Compact` operation succeeds.
-  - The failed `Compact` operation will still be retried in 1/10 of interval.
-  - e.g. `10h` compact interval will be retried in 1-hour on failure.
-  - Document...
 - Prevent [overflow by large `TTL` values for `Lease` `Grant`](https://github.com/coreos/etcd/pull/9399).
   - `TTL` parameter to `Grant` request is unit of second.
   - Leases with too large `TTL` values exceeding `math.MaxInt64` [expire in unexpected ways](https://github.com/coreos/etcd/issues/9374).

Documentation/op-guide/maintenance.md

@@ -21,9 +21,9 @@ $ etcd --auto-compaction-retention=1
 
 [v3.0.0](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.0.md) and [v3.1.0](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.1.md) with `--auto-compaction-retention=10` run periodic compaction on v3 key-value store for every 10-hour. Compactor only supports periodic compaction. Compactor records latest revisions every 5-minute, until it reaches the first compaction period (e.g. 10-hour). In order to retain key-value history of last compaction period, it uses the last revision that was fetched before compaction period, from the revision records that were collected every 5-minute. When `--auto-compaction-retention=10`, compactor uses revision 100 for compact revision where revision 100 is the latest revision fetched from 10 hours ago. If compaction succeeds or requested revision has already been compacted, it resets period timer and starts over with new historical revision records (e.g. restart revision collect and compact for the next 10-hour period). If compaction fails, it retries in 5 minutes.
 
-[v3.2.0](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.2.md) compactor runs [every hour](https://github.com/coreos/etcd/pull/7875). Compactor only supports periodic compaction. Compactor continues to record latest revisions every 5-minute. For every hour, it uses the last revision that was fetched before compaction period, from the revision records that were collected every 5-minute. That is, for every hour, compactor discards historical data created before compaction period. The retention window of compaction period moves to next hour. For instance, when hourly writes are about 100 and `--auto-compaction-retention=10`, v3.1 compacts revision 1000, 2000, and 3000 for every 10-hour, while v3.2 compacts revision 1000, 1100, and 1200 for every 1-hour. If compaction succeeds or requested revision has already been compacted, it resets period timer and removes used compacted revision from historical revision records (e.g. start next revision collect and compaction from previously collected revisions). If compaction fails, it retries in 5 minutes.
+[v3.2.0](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.2.md) compactor runs [every hour](https://github.com/coreos/etcd/pull/7875). Compactor only supports periodic compaction. Compactor continues to record latest revisions every 5-minute. For every hour, it uses the last revision that was fetched before compaction period, from the revision records that were collected every 5-minute. That is, for every hour, compactor discards historical data created before compaction period. The retention window of compaction period moves to next hour. For instance, when hourly writes are 100 and `--auto-compaction-retention=10`, v3.1 compacts revision 1000, 2000, and 3000 for every 10-hour, while v3.2.x, v3.3.0, v3.3.1, and v3.3.2 compact revision 1000, 1100, and 1200 for every 1-hour. If compaction succeeds or requested revision has already been compacted, it resets period timer and removes used compacted revision from historical revision records (e.g. start next revision collect and compaction from previously collected revisions). If compaction fails, it retries in 5 minutes.
 
-[v3.3.0](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.3.md) periodic compactor keeps recording latest revisions for every 1/10 of given compaction period (e.g. 1-hour when `--auto-compaction-mode=periodic --auto-compaction-retention=10h`). For every 1/10 of given compaction period, compactor uses the last revision that was fetched before compaction period, to discard historical data. The retention window of compaction period moves for every 1/10 of given compaction period. For instance, when hourly writes are about 100 and `--auto-compaction-retention=10`, v3.1 compacts revision 1000, 2000, and 3000 for every 10-hour, while v3.2 and v3.3 compacts revision 1000, 1100, and 1200 for every 1-hour. Futhermore, when writes per minute are about 1000, v3.3 with `--auto-compaction-mode=periodic --auto-compaction-retention=30m` can compact revision 30000, 33000, and 36000, for every 3-minute with more finer granularity. When `--auto-compaction-retention=10h`, etcd first waits 10-hour for the first compaction, and then does compaction every hour (1/10 of 10-hour) afterwards like this:
+[v3.3.0](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.3.md), [v3.3.1](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.3.md), and [v3.3.2](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.3.md) periodic compactor keeps recording latest revisions for every 1/10 of given compaction period (e.g. 1-hour when `--auto-compaction-mode=periodic --auto-compaction-retention=10h`). For every 1/10 of given compaction period, compactor uses the last revision that was fetched before compaction period, to discard historical data. The retention window of compaction period moves for every 1/10 of given compaction period. For instance, when hourly writes are 100 and `--auto-compaction-retention=10`, v3.1 compacts revision 1000, 2000, and 3000 for every 10-hour, while v3.2.x, v3.3.0, v3.3.1, and v3.3.2 compact revision 1000, 1100, and 1200 for every 1-hour. Futhermore, when writes per minute are about 1000, v3.3 with `--auto-compaction-mode=periodic --auto-compaction-retention=30m` can compact revision 30000, 33000, and 36000, for every 3-minute with more finer granularity. When `--auto-compaction-retention=10h`, etcd first waits 10-hour for the first compaction, and then does compaction every hour (1/10 of 10-hour) afterwards like this:
 
 ```
 0Hr  (rev = 1)
@@ -39,6 +39,8 @@ $ etcd --auto-compaction-retention=1
 
 Whether compaction succeeds or not, this process repeats for every 1/10 of given compaction period. If compaction succeeds, it just removes compacted revision from historical revision records.
 
+[v3.3.3](https://github.com/coreos/etcd/blob/master/CHANGELOG-3.3.md) periodic compactor keeps recording latest revisions for every compaction period when given period is less than 1-hour, or for every 1-hour when given compaction period is greater than 1-hour (e.g. 1-hour when `--auto-compaction-mode=periodic --auto-compaction-retention=24h`). For every compaction period or 1-hour, compactor uses the last revision that was fetched before compaction period, to discard historical data. The retention window of compaction period moves for every given compaction period or hour. For instance, when hourly writes are 100 and `--auto-compaction-mode=periodic --auto-compaction-retention=24h`, v3.2.x, v3.3.0, v3.3.1, and v3.3.2 compact revision 2400, 2640, and 2880 for every 2.4-hour, while v3.3.3 or later compacts revision 2400, 2500, 2600 for every 1-hour. Futhermore, when `--auto-compaction-mode=periodic --auto-compaction-retention=30m` and writes per minute are about 1000, v3.3.0, v3.3.1, and v3.3.2 compact revision 30000, 33000, and 36000, for every 3-minute, while v3.3.3 or later compacts revision 30000, 60000, and 90000.
+
 An `etcdctl` initiated compaction works as follows:
 
 ```sh