Commit Graph

772 Commits

Author SHA1 Message Date
2c3b3c093d rpc: do not leak grpc state change logger goroutine 2019-03-15 16:18:01 +01:00
ab3e783168 rpc: treat protocol handshake errors as permanent
treat handshake errors as permanent on the client

The issue was observed by 100% CPU usage due to lack ofrate-limiting in
dataconn.ReqPing retries=> safeguard that
2019-03-15 16:18:01 +01:00
7584c66bdb pruner: remove retry handling + fix early give-up
Retry handling is broken since the gRPC changes (wrong error classification).
Will come back at some point, hopefully by merging the replication
driver retry infrastructure.

However, the simpler architecture allows an easy fix for the problem
that the pruner practically gave up on the first error it encountered.

fixes #123
2019-03-13 21:04:39 +01:00
d78d20e2d0 pruner: skip placeholders + FSes without correspondents on source
fixes #126
2019-03-13 20:42:37 +01:00
b85ec52387 rpc/ctrl: nicer perr info debug log messages 2019-03-13 19:20:04 +01:00
edcd258cc9 replication: more elaborate messages for Conflict errors 2019-03-13 18:46:04 +01:00
d5250bbf51 client/status: fix wrap for multiline strings with leading space 2019-03-13 18:46:04 +01:00
d50e553ebb handle changes to placeholder state correctly
We assumed that `zfs recv -F FS` would basically replace FS inplace, leaving its children untouched.
That is in fact not the case, it only works if `zfs send -R` is set, which we don't do.

Thus, implement the required functionality manually.

This solves a `zfs recv` error that would occur when a filesystem previously created as placeholder on the receiving side becomes a non-placeholder filesystem (likely due to config change on the sending side):

  zfs send pool1/foo@1 | zfs recv -F pool1/bar
  cannot receive new filesystem stream:
  destination has snapshots (eg. pool1/bar)
  must destroy them to overwrite it
2019-03-13 18:46:04 +01:00
1eb0f12a61 replication: add diff test case 2019-03-13 18:45:40 +01:00
8129ed91f1 zfs + replication: migrate dead zfs/diff_test.go to replication/logic/diff
(and remove the dead code from package zfs)
2019-03-13 16:39:10 +01:00
c87759affe replication/driver: automatic retries on connectivity-related errors 2019-03-13 15:00:40 +01:00
07b43bffa4 replication: refactor driving logic (no more explicit state machine) 2019-03-13 15:00:40 +01:00
0230c6321f rpc/dataconn: microbenchmark 2019-03-13 13:57:21 +01:00
796c5ad42d rpc rewrite: control RPCs using gRPC + separate RPC for data transfer
transport/ssh: update go-netssh to new version
    => supports CloseWrite and Deadlines
    => build: require Go 1.11 (netssh requires it)
2019-03-13 13:53:48 +01:00
d281fb00e3 socketpair: directly export *net.UnixConn (and add test for that behavior) 2019-03-13 11:36:34 +01:00
76a6c623f3 tlsconf and transport/tls: support NSS-formatted keylog file for debugging
... via env variable
2019-03-13 00:28:38 +01:00
25c974f0b5 envconst: support for int64 2019-03-13 00:07:33 +01:00
ea719f5b5a build: use 'git describe --always' to determine ZREPL_VERSION 2019-03-13 00:07:33 +01:00
3105fa4ff8 build: use dep's required feature for dev tools 2019-03-12 23:43:39 +01:00
f724480c7b Add documentation regarding using a certificate chain 2019-01-22 10:09:24 -08:00
5c5e8c0baf Documentation changes mostly as requested 2019-01-22 16:46:34 +01:00
bb5278fe9b Permit peers to provide a cert chain (multiple certs). fixes #103 2019-01-09 10:10:37 -08:00
38b0bd76f5 build: just use go {test,vet} ./... for targets vet, test and generate 2018-12-11 22:00:03 +01:00
c1aab0bee9 config: update yaml-config and use zeropositive constraint for timeouts 2018-12-11 21:54:36 +01:00
ef3283638a logger: add stderrlogger (sometimes useful) 2018-12-11 21:24:54 +01:00
68b62a5c00 tlsconf: clear handshake deadline after completed handshake 2018-12-11 21:24:26 +01:00
7a75a4d384 util/iocommand: timeout kill on close + other hardening 2018-12-11 21:06:54 +01:00
1aae7b222f docs: fix confusing description of the role of client identity for sink jobs 2018-12-01 15:19:59 +01:00
3535b251ab freeze Go build dependencies in Gopkg.lock
* use pseudo-depdencies in build/build.go to convince dep
* update Travis, Dockerfile and Docs
* build.Dockerfile image now contains the Go build dependencies
* => faster builds
* bump pdu file after protoc update

fixes #106
2018-12-01 14:36:40 +01:00
707f070a3c build: fix dirty detection at the end of release build
was using Bashisms
2018-12-01 14:36:40 +01:00
160a3b6d32 more gofmt, drop snapjob.go_prefmt after it was accidentally added 2018-11-21 22:14:43 +01:00
d977796f18 Add SnapJob docs 2018-11-21 16:59:46 +01:00
3cef76d463 Refactor snapJob() to snapJobFromConfig() 2018-11-21 14:37:03 +01:00
e9564a7e5c Inlined a couple legacy leftover functions from the mode copypasta 2018-11-21 14:35:40 +01:00
b79ad3ddc3 Honour PruneKeepNotReplicated.KeepSnashotAtCursor in SinglePrunerFactory 2018-11-21 14:17:38 +01:00
d0f898751f Gofmt snapjob.go 2018-11-21 14:02:21 +01:00
22d9830baa Fix prometheus with multiple jobs 2018-11-21 04:26:03 +01:00
c4e23862cd Added status view for SnapJob. 2018-11-21 04:06:13 +01:00
e10dc129de Make getPruner() private 2018-11-21 03:39:03 +01:00
dd11fc96db Touchups in job.go 2018-11-21 03:27:39 +01:00
7de3c0a09a Removed the references to a pruning 'side' in the singlepruner logging code and the snapjob prometheus thing. 2018-11-21 02:52:33 +01:00
141e49727c Missed a last reference to tasks 2018-11-21 02:51:23 +01:00
442d61918b remove most of the watchdog machinery 2018-11-21 02:42:13 +01:00
58dcc07430 Added SnapJobStatus 2018-11-21 02:08:39 +01:00
19d0916e34 remove snapMode, rename snap_ActiveSide to SnapJob 2018-11-21 01:54:56 +01:00
1265cc7934 pruned unused lines and comments ;) 2018-11-21 01:34:50 +01:00
3d2688e959 Ugly but working inital snapjob implementation 2018-11-20 19:30:15 +01:00
7ab51fad0d zfs: add 'received' property source, handle 'any' source correctly and use 'any' for placeholder FS detection
we want was first noticed in zfs 0.8rc1

Upstream doc PR: https://github.com/zfsonlinux/zfs/pull/8134
2018-11-16 13:07:13 +01:00
3472145df6 pruner + proto change: better handling of missing replication cursor
- don't treat missing replication cursor as an error in protocol
- treat it as a per-fs planning error instead
2018-11-16 12:21:54 +01:00
5e1ea21f85 pruning: add 'Negate' option to KeepRegex and expose it in config 2018-11-16 12:21:54 +01:00