6a33f0ffd5
clientv3: make balancer respect FastFail
...
The simpleBalancer.Get() blocks grpc.Invoke() even when the Invoke() is called
with the FailFast option. Therefore currently any requests with the
FastFail option actually doesn't fail fast. They get blocked when there is
no endpoints available.
Get() method needs to respect the BlockingWait option when
picks up an endpoint address from the list and fail immediately when the option is
enabled and no endpoint is available.
2016-10-17 14:11:51 -07:00
7d50dc06a2
clientv3/naming: support resolving to multiple hosts
...
Previous implementation watches a single key so there's no way
to have separate hosts associate with separate keys for a single
grpc target. Instead, accept all keys on a prefix.
Also fixes first the Next() to read current name data from etcd instead
of waiting for the next event on a synced watcher.
2016-10-12 11:27:22 -07:00
084c407a8d
clientv3: drop Config.Logger field
...
Fix https://github.com/coreos/etcd/issues/6603 .
Instead adds 'SetLogger' to set global logger interface
to avoid unnecessary logger updates.
2016-10-11 16:38:32 -07:00
4a07bbec59
clientv3: implement new grpc.Balancer interface
2016-10-10 11:18:29 -07:00
54c252ee63
clientv3/kv_test: Fix quota test
...
Updates TestKVPutError. Change the quota to work with systems
that have a 64 KiB page size. Increase the db sync wait time to
one second. Also, add some comments for the hard coded value.
Signed-off-by: Geoff Levand <geoff@infradead.org>
2016-10-05 16:41:06 -07:00
b980ab0c67
Merge pull request #6582 from heyitsanthony/fix-cancel-close
...
clientv3: only return closing error to watcher if context is not canceled
2016-10-05 13:37:29 -07:00
06d5cf2d52
clientv3: fix race on watch initial revision
...
The initial revision was being updated in the substream goroutine defer;
this was racing with the resume path fetching the initial revision when
the substream closes during resume. Instead, update the initial revision
whenever the substream processes a new watch response. Since the substream
cannot receive a watch response while it is resuming, the write to the
initial revision is ordered to always happen after the resume read.
Fixes #6586
2016-10-05 09:36:06 -07:00
e285f599e2
clientv3: only return closing error to watcher if context is not canceled
...
Fixes #6503
2016-10-04 16:09:50 -07:00
8e1c989ec3
integration: test a canceled watch won't return a closing error
2016-10-04 14:47:40 -07:00
5b50658118
clientv3: simplify watch synchronization
...
Was more complicated than it needed to be and didn't really work in the
first place. Restructured watcher registation to use a queue.
2016-10-03 16:56:14 -07:00
9ce398f8a6
integration: test canceling watchers when disconnected
2016-10-03 16:56:14 -07:00
29dd3cf5bd
Revert "clientv3/integration: add TestDialWithHTTPS"
...
This reverts commit a96a28d603
2016-10-03 13:05:08 -07:00
a96a28d603
clientv3/integration: add TestDialWithHTTPS
2016-10-03 02:16:07 -07:00
f45542394b
clientv3: handle 'https' scheme in endpoint
2016-10-03 01:03:28 -07:00
bf2581390d
clientv3: make IsProgressNotify() false on compact event and closed channel
...
Fixes #6549
2016-09-28 16:49:39 -07:00
27c500d8d0
Merge pull request #6487 from heyitsanthony/watch-stress
...
clientv3: process closed watcherStreams in watcherGrpcStream run loop
2016-09-21 13:55:25 -07:00
a32518006c
clientv3: process closed watcherStreams in watcherGrpcStream run loop
...
Was racing with Watch() when closing the grpc stream on no watchers.
Fixes #6476
2016-09-21 13:28:00 -07:00
deef16b376
integration: test client watchers with overlapped context cancels
2016-09-21 09:40:24 -07:00
cdb1e34799
clientv3: add 'Sync' method
2016-09-21 09:10:25 -07:00
69c7ea0b4a
Merge pull request #6473 from heyitsanthony/watchreconn-putretry
...
integration: l-read before Put in TestWatchReconnRequest
2016-09-19 14:52:26 -07:00
0fb2cab221
integration: l-read before Put in TestWatchReconnRequest
...
TestWatchReconnRequest occasionally triggers elections because it spins on
drop connections, eating up CPU. In case there's an election, submit an
l-read to wait for the cluster to settle down.
Fixes #6314
2016-09-19 14:14:32 -07:00
f11b35eb71
clientv3/integration: test 'SetEndpoints'
2016-09-20 04:36:14 +09:00
b9d18d4ac9
clientv3: add 'SetEndpoints' method
2016-09-20 04:36:01 +09:00
e77baa3dcb
Merge pull request #6424 from heyitsanthony/v3api-createminmax
...
etcdserver: range queries with min/max create revision
2016-09-14 19:10:52 -07:00
059f419ac5
Merge pull request #6429 from xiang90/fix_balancer
...
clientv3: balancer panics when call up after close
2016-09-14 19:42:24 -05:00
9b1fe45853
concurrency: use create max revision for locks and elections
2016-09-14 17:03:33 -07:00
004a5f0dbc
clientv3: balancer panics when call up after close
...
Fix the issue by adding a simple guard varable.
2016-09-15 07:43:42 +08:00
5bd251a6fa
clientv3: WithMinCreateRev, WithMaxCreateRev
2016-09-14 15:31:45 -07:00
4211c0b7af
etcdctl, clientv3: improve printing of role get for prefix permission
...
This commit improves printing of role get command for prefix
permission. If a range permission corresponds to a prefix permission,
it is explicitly printed for a user. Below is an example of the new
printing:
$ ETCDCTL_API=3 bin/etcdctl --user root:p role get r1
Role r1
KV Read:
[/dir/, /dir0) (prefix /dir/)
[k1, k5)
KV Write:
[/dir/, /dir0) (prefix /dir/)
[k1, k5)
2016-09-14 16:10:32 +09:00
e6a7f25065
Merge pull request #6411 from heyitsanthony/v3api-minmaxmod
...
etcdserver: Range with min/max mod revision
2016-09-13 05:54:58 -05:00
56084a7cc8
clientv3: return error from response when possible
2016-09-13 11:18:21 +08:00
b0481ba858
clientv3: WithMinModRev and WithMaxModRev
2016-09-12 19:44:14 -07:00
03b9d6f24c
*: separate 'capnslog' log level setting
2016-09-10 20:26:51 +09:00
4c08f6767c
clientv3: add lease.TimeToLive + tests
2016-09-09 08:18:45 +09:00
1aec483e42
clientv3: use correct context in toErr (lease)
2016-09-08 10:58:11 +09:00
ad318ee891
clientv3: drain buffered WatchResponses before resuming
...
Otherwise, the watcherStream can receive WatchResponses in the
middle of a resume, corrupting the stream.
Fixes #6364
2016-09-06 17:15:39 -07:00
bc5d7bbe03
auth, e2e, clientv3: the root role should be granted access to every key
...
This commit changes the semantics of the root role. The role should be
able to access to every key.
Partially fixes https://github.com/coreos/etcd/issues/6355
2016-09-06 16:10:28 +09:00
7b2f769643
clientv3: only resume watcher if error is non-halting
2016-09-01 15:22:35 -07:00
9497e9678c
clientv3/concurrency: allow election on prefixes of keys.
...
After winning an election or obtaining a lock, we
auto-append a slash after the provided key prefix.
This avoids the previous deadlock due to waiting
on the wrong key.
Fixes #6278
2016-08-29 18:34:14 -07:00
5e963608b7
clientv3: do not treat Internal codes as halting
...
Fixes #6277
2016-08-28 20:20:22 -07:00
3552420dfd
clientv3: set failfast=false on read-only txns
2016-08-28 19:40:38 -07:00
a50c707050
clientv3/integration: wait for two request timeouts in txn tests
...
Read only txns and Get may timeout once if the leader is lost.
2016-08-26 10:04:10 -07:00
267063efd0
clientv3: use grpc codes to translate raw grpc errors
2016-08-26 09:22:09 -07:00
f2b7c501cc
clientv3/integration: fix TestKVPutStoppedServerAndClose
2016-08-24 13:57:27 -07:00
be38c50567
clientv3: specify watch progress notify interval
...
For watch request
2016-08-22 15:44:59 -07:00
8d7703528a
Merge pull request #5845 from heyitsanthony/clientv3-ignore-dead-eps
...
clientv3: respect up/down notifications from grpc
2016-08-16 11:56:03 -07:00
3eadf964f4
clientv3: use failfast and retry wrappers for at-most-once rpcs
2016-08-16 10:49:50 -07:00
46765ad79c
clientv3: respect up/down notifications from grpc
...
Fixes #5842
2016-08-16 09:49:36 -07:00
feaff17259
session: remove session manager and add ttl
2016-08-15 14:12:25 -07:00
429d5ab20b
clientv3: only block on New() when DialTimeout > 0
...
Fixes #6162
2016-08-12 10:33:11 -07:00
Iwasaki Yudai