Commit Graph

147 Commits

Author SHA1 Message Date
Christian Schwarz
f9d24d15ed move wakup mechanism into separate package 2018-10-12 12:44:40 +02:00
Christian Schwarz
1fb59c953a implement transport protocol handshake (even before streamrpc handshake) 2018-10-11 21:21:46 +02:00
Christian Schwarz
be962998ba move serve and connecter into transports package 2018-10-11 21:21:46 +02:00
Christian Schwarz
a97684923a refactor: socketpair into utils package (useful elsewhere) 2018-10-11 21:17:43 +02:00
Christian Schwarz
125b561df3 rename root_dataset to root_fs for receiving-side jobs 2018-10-11 18:03:18 +02:00
Christian Schwarz
4e16952ad9 snapshotting: support 'periodic' and 'manual' mode
1. Change config format to support multiple types
   of snapshotting modes.
2. Implement a hacky way to support periodic or completely
   manual snaphots.

In manual mode, the user has to trigger replication using the wakeup
mechanism after they took snapshots using their own tooling.

As indicated by the comment, a more general solution would be desirable,
but we want to get the release out and 'manual' mode is a feature that
some people requested...
2018-10-11 15:59:23 +02:00
Christian Schwarz
01668a989e transport local: named listeners + struct renaming 2018-10-11 13:06:47 +02:00
Christian Schwarz
75e42fd860 pruner: implement Report method + display in status command 2018-09-24 19:25:40 +02:00
Christian Schwarz
75ba5874a5 active side: track activities in Run() as atomically updated member 2018-09-24 19:23:53 +02:00
Christian Schwarz
1ce0c69e4f implement local replication using new local transport
The new local transport uses socketpair() and a switchboard based on
client identities.
The special local job type is gone, which is good since it does not fit
into the 'Active/Passive side ' + 'mode' concept used to implement the
duality of push/sink | pull/source.
2018-09-24 14:43:53 +02:00
Christian Schwarz
f3e8eda04d fixup 4e04f8d3d2: snapper with separate stopped state for clean shutdown
would tight loop in ErrorWait
2018-09-24 14:40:47 +02:00
Christian Schwarz
4e04f8d3d2 snapper: make error mode an error wait mode
Just because taking one snapshot fails does not mean snapper needs to
stop for all others.
Since users are advised to monitor error logs, snapshot-taking errors
can still be addressed.
The ErrorWait mode allows a potential future Report / Status command to
distinguish normal waits from error waits.
2018-09-24 12:36:10 +02:00
Christian Schwarz
d04b9713c4 implement pull + sink modes for active and passive side 2018-09-24 12:36:10 +02:00
Christian Schwarz
ffe33aff3d fix pruner: protobuf one-ofs require non-zero value, even if no public fields 2018-09-24 12:36:10 +02:00
Christian Schwarz
e3be120d88 refactor push + source into active + passive 'sides' with push and source 'modes' 2018-09-24 12:36:10 +02:00
Christian Schwarz
9446b51a1f status: infra for reporting jobs instead of just replication.Report 2018-09-23 21:11:33 +02:00
Christian Schwarz
4a6160baf3 update to streamrpc 0.4 & adjust config (not breaking) 2018-09-23 20:28:30 +02:00
Christian Schwarz
7f9eb62640 sink: concurrent connection handling 2018-09-18 22:44:00 +02:00
Christian Schwarz
6c3f442f13 daemon control / jsonclient: fix connection leak due to open request body
Also:
- Defensive measures in control http server (1s timeouts)
(prevent the leak, even if request body is not closed)
- Add prometheus metrics to track control socket latencies
(were used for debugging)
2018-09-13 12:44:46 +02:00
Christian Schwarz
fa47667f31 bring back prometheus metrics, with new metrics for replication state machine 2018-09-07 22:22:34 -07:00
Christian Schwarz
7836ea36fc serve TLS: validate client CNs against whitelist in config file 2018-09-06 13:34:39 -07:00
Christian Schwarz
0f75677e59 daemon/pruner: fix exercise (don't call it test) 2018-09-05 21:47:44 -07:00
Christian Schwarz
2c25f28972 simplify mapping & filtering in endpoints (re-rooting only) 2018-09-05 19:51:06 -07:00
Christian Schwarz
975fdee217 replication & pruning: ditch replicated-property, use bookmark as cursor instead
A bookmark with a well-known name is used to track which version was
last successfully received by the receiver.
The createtxg that can be retrieved from the bookmark using `zfs get` is
used to set the Replicated attribute of each snap on the sender:
If the snap's CreateTXG > the cursor's, it is not yet replicated,
otherwise it has been.

There is an optional config option to change the behvior to
`CreateTXG >= the cursor's`, and the implementation defaults to that.

The reason: While things work just fine with `CreateTXG > the cursor's`,
ZFS does not provide size estimates in a `zfs send` dry run
(see acd2418).
However, to enable the use case of keeping the snapshot only around for
the replication, the config flag exists.
2018-09-05 19:51:06 -07:00
Christian Schwarz
9eca269ad8 fixup 308e5e35fb: remove fprintf debug output 2018-09-05 08:35:31 -07:00
Christian Schwarz
c21222ef13 serve/tls: use handshake timeout from config 2018-09-05 08:32:59 -07:00
Christian Schwarz
adab06405b make go vet happy 2018-09-04 17:25:10 -07:00
Christian Schwarz
bfc631f6a6 fix broken pruner exercise (don't call it test...) 2018-09-04 17:19:59 -07:00
Christian Schwarz
308e5e35fb Multi-client servers + bring back stdinserver support 2018-09-04 16:43:55 -07:00
Christian Schwarz
e161347e47 Implement periodic snapshotting. 2018-09-04 16:43:55 -07:00
Christian Schwarz
754b253043 config: no-field for replication anymore
It's closer to the original config and we don't want users to specify
'filesystems' and similar multiple times in a single job definition.
2018-09-04 14:44:45 -07:00
Christian Schwarz
0c4a3f8dc4 pruning/history: properly communicate via rpc if snapshot does not exist 2018-09-04 14:01:48 -07:00
Christian Schwarz
8799108b55 fixup b95e983d0d: prunerFactory: fix duplicate logger fields 2018-09-03 13:19:56 -07:00
Christian Schwarz
03f9f81cb5 fixup 3d8e552c6a: validate streamrpc config in factory constructors 2018-09-03 13:17:53 -07:00
Christian Schwarz
3d8e552c6a streamrpc 0.3 + config from daemon/config 2018-09-02 15:46:42 -07:00
Christian Schwarz
d55a271ac7 WIP adopt updated yaml-config with 'fromdefaults' struct tag 2018-09-02 15:46:03 -07:00
Christian Schwarz
b95e983d0d bump go-streamrpc to 0.2, cleanup logging
logging should be user-friendly in INFO mode
2018-09-02 15:45:18 -07:00
Anton Schirg
32391adf4f build pruner in factory and check prune rules 2018-08-30 19:20:14 +02:00
Anton Schirg
c0a3e1f121 wrap error in buildJob with job name 2018-08-30 17:40:02 +02:00
Christian Schwarz
1690339440 colorized stdout logger if stdout is tty 2018-08-30 13:33:28 +02:00
Anton Schirg
6cedd0a2e8 add status command 2018-08-30 12:54:30 +02:00
Christian Schwarz
7dd49b835a finish pruning implementation in push job 2018-08-30 11:52:05 +02:00
Christian Schwarz
12dd240b5f fixup pruner 2018-08-30 11:49:06 +02:00
Christian Schwarz
a2aa8e7bd7 finish pruner implementation 2018-08-29 19:00:45 +02:00
Christian Schwarz
c69ebd3806 WIP rewrite the daemon
cmd subdir does not build on purpose, it's only left in tree to grab old
code and move it to github.com/zrepl/zrepl/daemon
2018-08-27 22:22:44 +02:00
Christian Schwarz
df6e1bc64d privatize pprofServer 2018-08-27 19:13:35 +02:00
Christian Schwarz
89dc267780 start implementing new daemon in package daemon 2018-08-27 19:10:55 +02:00