ffa87f9 | Yicheng Qin | 25 August 2015, 03:06:13 UTC | storage: fix the comment in generation.walk | 26 August 2015, 19:53:30 UTC |
76db974 | Yicheng Qin | 26 August 2015, 05:45:10 UTC | Merge pull request #3377 from yichengq/tls-info-string pkg/transport: print ClientCertAuth in TLSInfo.String() | 26 August 2015, 05:45:10 UTC |
45bb880 | Yicheng Qin | 26 August 2015, 05:38:36 UTC | Merge pull request #3378 from yichengq/set-late etcdmain: check error before assigning peer transport | 26 August 2015, 05:38:36 UTC |
58455a2 | Yicheng Qin | 26 August 2015, 04:59:56 UTC | etcdmain: check error before assigning peer transport Or it may panic when new transport fails, e.g., TLS info is invalid. | 26 August 2015, 05:04:26 UTC |
57e8846 | Yicheng Qin | 26 August 2015, 04:53:52 UTC | pkg/transport: print ClientCertAuth in TLSInfo.String() It is good to print it in debug output: ``` 21:56:12 etcd1 | 2015-08-25 21:56:12.162406 I | etcdmain: peerTLS: cert = certs/etcd1.pem, key = certs/etcd1-key.pem, ca = , trusted-ca = certs/ca.pem, client-cert-auth = true ``` | 26 August 2015, 04:53:52 UTC |
6250fed | Yicheng Qin | 26 August 2015, 03:08:19 UTC | Merge pull request #3096 from philips/tls-info-debug pkg/transport: include debug output for trusted-ca | 26 August 2015, 03:08:19 UTC |
008f988 | Xiang Li | 25 August 2015, 21:48:35 UTC | Merge pull request #3375 from xiang90/doc doc: add evn variable name to configuration.md | 25 August 2015, 21:48:35 UTC |
2b58da1 | Yicheng Qin | 25 August 2015, 21:48:00 UTC | Merge pull request #3374 from yichengq/gomaxprocs etcdmain: change default GOMAXPROCS when compiling in go1.5 | 25 August 2015, 21:48:00 UTC |
35a0459 | Xiang Li | 25 August 2015, 21:35:15 UTC | doc: add evn variable name to configuration.md | 25 August 2015, 21:35:15 UTC |
32ab3f6 | Xiang Li | 25 August 2015, 21:04:30 UTC | Merge pull request #3372 from xiang90/doc improve clustering.md doc | 25 August 2015, 21:04:30 UTC |
c30c858 | Xiang Li | 25 August 2015, 18:32:46 UTC | doc: add explanation for client urls | 25 August 2015, 20:46:27 UTC |
2ac9a32 | Yicheng Qin | 25 August 2015, 19:08:52 UTC | etcdmain: stop setting GOMAXPROCS explicitly We always want to use GOMAXPROCS() as the way go parses it. When in go1.4, we want to expose GOMAXPROCS value, so we set GOMAXPROCS explicitly as the way go 1.4 does and print it out. But it becomes a problem when go 1.5 changes the way to set GOMAXPROCS. Fix the problem by stop setting GOMAXPROCS and get its value directly. Due to this change, it sets default GOMAXPROCS to the number of CPUs available when compiling in go 1.5, which matches how go 1.5 works: https://docs.google.com/document/d/1At2Ls5_fhJQ59kDK2DFVhFu3g5mATSXqqV5QrxinasI/edit This is a behavior change in etcd 2.2. | 25 August 2015, 20:38:16 UTC |
a4285ef | Yicheng Qin | 25 August 2015, 20:12:48 UTC | Merge pull request #3367 from MSamman/master etcdserver: handle malformed basic auth | 25 August 2015, 20:12:48 UTC |
e2e002f | Mohammad Samman | 24 August 2015, 18:10:19 UTC | etcdserver: handle malformed basic auth return insufficient credentials if basic auth header is malformed Fixes #3280 | 25 August 2015, 19:37:24 UTC |
7bd558b | Xiang Li | 25 August 2015, 19:17:06 UTC | Merge pull request #3373 from ecnahc515/add_report_bugs_contributing Contributing: Link to reporting bugs doc | 25 August 2015, 19:17:06 UTC |
ad84334 | Chance Zibolski | 25 August 2015, 19:13:10 UTC | Contributing: Link to reporting bugs doc | 25 August 2015, 19:15:03 UTC |
f56c545 | Xiang Li | 25 August 2015, 18:22:08 UTC | doc: mention reconfiguration design in clustering.md | 25 August 2015, 18:22:08 UTC |
986f354 | Xiang Li | 25 August 2015, 18:17:14 UTC | Merge pull request #3371 from xiang90/bolt Godeps: update bolt dependency | 25 August 2015, 18:17:14 UTC |
e8f40b0 | Xiang Li | 25 August 2015, 17:57:23 UTC | storage/backend: add commitAndStop After the upgrade of boltdb, db.Close waits for all txn to finish. CommitAndStop commits the current txn and stop creating new ones. | 25 August 2015, 17:57:25 UTC |
8738a88 | Xiang Li | 25 August 2015, 17:39:29 UTC | Godeps: update bolt dependency | 25 August 2015, 17:39:29 UTC |
2d06f6b | Yicheng Qin | 25 August 2015, 16:26:46 UTC | Merge pull request #3362 from yichengq/rafthttp-cancel rafthttp: always cancel in-flight request when stop streamReader | 25 August 2015, 16:26:46 UTC |
61a75b3 | Yicheng Qin | 24 August 2015, 17:33:21 UTC | rafthttp: always cancel in-flight request when pipeline.send This fits the way for go1.5 to cancel request. | 25 August 2015, 16:07:49 UTC |
27b9963 | Yicheng Qin | 24 August 2015, 17:42:28 UTC | client: always cancel in-flight request when do request This fits the way for go1.5 to cancel request. | 25 August 2015, 16:04:58 UTC |
ece39c9 | Yicheng Qin | 24 August 2015, 17:51:17 UTC | proxy: always cancel in-flight request This fits the way for go1.5 to cancel request. | 25 August 2015, 15:59:59 UTC |
6fc6386 | Yicheng Qin | 25 August 2015, 15:52:48 UTC | rafthttp: return err if stopped before setting cancel in dial() The original workflow may fail to cancel if stop() cancels the finished request just before dial() assigning a new cancel. This commit checks streamReader status before setting cancel to avoid this problem. It is tested at travis for 300 times. go 1.5 always works well, while go 1.4 fails to stop once. | 25 August 2015, 15:59:12 UTC |
fc95ec0 | Yicheng Qin | 24 August 2015, 14:28:36 UTC | rafthttp: always cancel in-flight request when stop streamReader This problem is totally fixed at 1.5. go1.5 adds a Request.Cancel channel, which allows for "race free" cancellation (https://github.com/golang/go/commit/8b4278ffb75e79c277bfa90c5e473bfad9f7c1bd). Our implementation relies on it to always cancel in-flight request. | 25 August 2015, 15:54:13 UTC |
0132b09 | Yicheng Qin | 25 August 2015, 14:58:30 UTC | Merge pull request #3360 from yichengq/bench-3 *: add initial read benchmark for etcd v3 | 25 August 2015, 14:58:30 UTC |
3632a1b | Yicheng Qin | 23 August 2015, 06:02:22 UTC | *: add initial read benchmark for etcd v3 It includes the initial read benchmark for etcd v3. This is the first step to give some rough thoughts. I haven't digged deeper to answer some questions, including why its performance is not better than HTTP + json, why one put will cause performance downgrade. | 25 August 2015, 14:50:18 UTC |
e3ef1d3 | Xiang Li | 24 August 2015, 18:22:29 UTC | Merge pull request #3366 from xiang90/v3_proto update v3 proto and doc | 24 August 2015, 18:22:29 UTC |
0cb45ae | Xiang Li | 24 August 2015, 17:58:05 UTC | rfc: update v3 proto | 24 August 2015, 18:00:51 UTC |
1cccbb5 | Xiang Li | 24 August 2015, 17:52:54 UTC | etcdserverpb: add comments for compaction | 24 August 2015, 17:52:54 UTC |
3a60d49 | Xiang Li | 24 August 2015, 17:42:16 UTC | storagepb: fix comment location | 24 August 2015, 17:42:16 UTC |
4a5b944 | Xiang Li | 24 August 2015, 17:40:05 UTC | etcdserverpb: update comment for txn request | 24 August 2015, 17:40:05 UTC |
98ceb3c | Xiang Li | 24 August 2015, 17:33:20 UTC | etcdserverpb: add more field into rangeResponse | 24 August 2015, 17:33:20 UTC |
c7f10ed | Yicheng Qin | 24 August 2015, 16:44:13 UTC | Merge pull request #3361 from yichengq/no-log integration: only print critical log | 24 August 2015, 16:44:13 UTC |
3702be4 | Yicheng Qin | 24 August 2015, 04:16:18 UTC | integration: only print critical log This limits the logs printed out in integration test, so it will not have log flood and help us read fatal log in travis. | 24 August 2015, 04:22:21 UTC |
514c437 | Yicheng Qin | 24 August 2015, 04:12:36 UTC | Merge pull request #3359 from yichengq/storage-test functional tests for storage package and some related fixes | 24 August 2015, 04:12:36 UTC |
1e2b0ac | Yicheng Qin | 23 August 2015, 00:40:11 UTC | test: activate test for storage package | 24 August 2015, 03:59:06 UTC |
9c0c314 | Yicheng Qin | 23 August 2015, 00:38:54 UTC | storage: add functional tests for the package It adds and reorganize tests to construct functional tests. | 24 August 2015, 03:59:06 UTC |
9960651 | Yicheng Qin | 23 August 2015, 00:30:59 UTC | storage: let range work in the process of txn range should work in the process of txn to help check the status during the txn. | 24 August 2015, 03:59:06 UTC |
6d97dca | Yicheng Qin | 22 August 2015, 23:47:19 UTC | storage: ensure that desired compaction is persisted It needs to persist the desired compaction, so it won't forget the compaction if it crashes later. | 24 August 2015, 03:59:06 UTC |
353f10c | Yicheng Qin | 22 August 2015, 23:44:39 UTC | storage: reject to compact on future rev Compaction on future rev is unreasonable. | 24 August 2015, 03:59:06 UTC |
47b243b | Yicheng Qin | 22 August 2015, 23:42:27 UTC | storage: let TxnDeleteRange return rev if no error If it doesn't return error, it should return valid rev. | 24 August 2015, 03:59:06 UTC |
62f7481 | Yicheng Qin | 22 August 2015, 23:39:29 UTC | storage: keyIndex.get returns err when key is tombstoned Before this commit, it will return wrong create index, mod index. It lets findGeneration return error when rev is at the gap of two generations. This leads to the change of compact() code. | 24 August 2015, 03:59:02 UTC |
3b2fa9f | Yicheng Qin | 23 August 2015, 00:57:12 UTC | storage: fix TestKeyIndexCompact It fails to pass before. | 24 August 2015, 00:22:49 UTC |
97b211c | Xiang Li | 22 August 2015, 17:29:29 UTC | Merge pull request #3357 from ccding/master go vet | 22 August 2015, 17:29:29 UTC |
c09b667 | Cong Ding | 22 August 2015, 17:07:53 UTC | *: fix go vet reported issues | 22 August 2015, 17:19:02 UTC |
044b23c | Xiang Li | 22 August 2015, 01:59:51 UTC | Merge pull request #3356 from xiang90/travis *: test gofmt with -s and fix reported issues | 22 August 2015, 01:59:51 UTC |
6b23a81 | Xiang Li | 22 August 2015, 01:52:16 UTC | *: test gofmt with -s and fix reported issues | 22 August 2015, 01:52:16 UTC |
301b7f5 | Yicheng Qin | 21 August 2015, 22:37:15 UTC | Merge pull request #3355 from yichengq/health-var etcdctl/cluster_health: set health var when checked healthy | 21 August 2015, 22:37:15 UTC |
2247558 | Yicheng Qin | 21 August 2015, 22:27:35 UTC | etcdctl/cluster_health: set health var when checked healthy This was a typo. | 21 August 2015, 22:27:35 UTC |
84b614c | Xiang Li | 21 August 2015, 21:49:00 UTC | Merge pull request #3342 from xiang90/travis travis: test for go 1.5 build | 21 August 2015, 21:49:00 UTC |
1dcc145 | Xiang Li | 21 August 2015, 21:36:29 UTC | client: fix test | 21 August 2015, 21:36:29 UTC |
8c0610d | Yicheng Qin | 21 August 2015, 19:38:38 UTC | Merge pull request #3352 from yichengq/fix-name-url fix that etcd fails to start if using both IP and hostname when discovery srv | 21 August 2015, 19:38:38 UTC |
3c1e6b5 | Yicheng Qin | 21 August 2015, 19:09:17 UTC | pkg/netutil: stop resolving in place It helps to copy out a and b, and not modify the original a and b. | 21 August 2015, 19:09:17 UTC |
1c33497 | Yicheng Qin | 21 August 2015, 19:01:00 UTC | pkg/netutil: not introduce empty url when converting It should not make slices with length and append elements at the same time. | 21 August 2015, 19:08:17 UTC |
7b871aa | Yicheng Qin | 21 August 2015, 18:58:37 UTC | pkg/netutil: not export resolve and urlsEqual functions They are only used in this package, so there is no need to public them. | 21 August 2015, 18:58:37 UTC |
b1192e5 | Yicheng Qin | 21 August 2015, 17:15:08 UTC | pkg/netutil: fix false negative comparison Sort the resolved URLs before DeepEqual, so it will not compare URLs that may be out of order due to resolution. | 21 August 2015, 17:15:08 UTC |
72462a7 | Yicheng Qin | 21 August 2015, 16:51:07 UTC | etcdserver: remove TODO to delete URLStringsEqual Discovery SRV supports to compare IP addresses with domain names, so we need URLStringsEqual function. | 21 August 2015, 16:52:17 UTC |
8ea3d15 | Yicheng Qin | 21 August 2015, 16:41:13 UTC | Revert "Revert "Treat URLs have same IP address as same"" This reverts commit 3153e635d5799f1ff8715fd4b7c07e0283820acd. Conflicts: etcdserver/config.go | 21 August 2015, 16:41:13 UTC |
07af0b3 | Xiang Li | 21 August 2015, 06:32:29 UTC | Merge pull request #3346 from xiang90/auth_skip etcdserver/auth: cache auth enable result | 21 August 2015, 06:32:29 UTC |
11a689d | Xiang Li | 20 August 2015, 23:13:27 UTC | etcdserver/auth: cache auth enable result | 21 August 2015, 06:05:00 UTC |
e8e507b | Xiang Li | 21 August 2015, 05:44:51 UTC | Merge pull request #3348 from xiang90/l use limited listener from golang | 21 August 2015, 05:44:51 UTC |
ff37cc4 | Xiang Li | 21 August 2015, 03:03:27 UTC | pkg/transport: remove home-grown limitedListener | 21 August 2015, 03:03:27 UTC |
9263435 | Xiang Li | 21 August 2015, 03:02:35 UTC | *: use limitedListener from golang | 21 August 2015, 03:02:35 UTC |
da9a12b | Xiang Li | 20 August 2015, 22:10:25 UTC | Merge pull request #3344 from xiang90/startup_version etcdmain: print out version information on startup | 20 August 2015, 22:10:25 UTC |
6b77c14 | Xiang Li | 20 August 2015, 21:48:53 UTC | etcdmain: print out version information on startup | 20 August 2015, 21:50:16 UTC |
31395d2 | Xiang Li | 20 August 2015, 18:39:41 UTC | travis: test for go 1.5 build | 20 August 2015, 18:39:41 UTC |
7cf9770 | Xiang Li | 20 August 2015, 18:38:03 UTC | Merge pull request #3340 from xiang90/fix_perallocate pkg/fileutil: treat not support error as nil error in preallocate | 20 August 2015, 18:38:03 UTC |
3ca5482 | Xiang Li | 20 August 2015, 18:02:56 UTC | pkg/fileutil: treat not support error as nil error in preallocate | 20 August 2015, 18:15:02 UTC |
4a6d6b0 | Xiang Li | 20 August 2015, 17:16:31 UTC | Merge pull request #3338 from spacejam/master Reversion->Revision | 20 August 2015, 17:16:31 UTC |
acd7a92 | Tyler Neely | 20 August 2015, 15:39:07 UTC | storage: reversion -> revision | 20 August 2015, 15:39:07 UTC |
e1dfcec | Yicheng Qin | 20 August 2015, 07:18:32 UTC | Merge pull request #3327 from yichengq/bench-2.2 docs/benchmarks: add benchmark result for 2.2 | 20 August 2015, 07:18:32 UTC |
807de81 | Yicheng Qin | 19 August 2015, 05:37:49 UTC | docs/benchmarks: add benchmark result for 2.2 And it analyzes the reason for performance changes. | 20 August 2015, 06:59:33 UTC |
795e962 | Yicheng Qin | 20 August 2015, 03:59:04 UTC | Merge pull request #3334 from mitake/snap-marsharing-prometheus snap: export durations of marsharing cost during snapshot save | 20 August 2015, 03:59:04 UTC |
7a6d336 | Hitoshi Mitake | 20 August 2015, 01:58:29 UTC | snap: export durations of marshalling cost during snapshot save Currently, total duration of snapshot saving is exported for prometheus. For more detailed analysis, this commit let etcd export durations of marshalling for prometheus. | 20 August 2015, 03:47:07 UTC |
46a2ae7 | Yicheng Qin | 19 August 2015, 23:51:39 UTC | hack/benchmark: add script for benchmark This is for etcd benchmark. | 20 August 2015, 03:37:27 UTC |
b0303e9 | Xiang Li | 20 August 2015, 00:15:52 UTC | Merge pull request #3323 from xiang90/cl_health etcdctl: use health endpoint to greatly simplify health checking | 20 August 2015, 00:15:52 UTC |
568d1c6 | Xiang Li | 18 August 2015, 22:04:41 UTC | etcdctl: use health endpoint to greatly simplify health checking | 19 August 2015, 18:47:08 UTC |
60387dc | Xiang Li | 19 August 2015, 18:08:05 UTC | Merge pull request #3320 from yichengq/doc-rtt docs: document how to set heartbeat interval and election timeout | 19 August 2015, 18:08:05 UTC |
28b61ac | Xiang Li | 19 August 2015, 00:18:08 UTC | Merge pull request #3324 from xiang90/raft_logging raft: downgrade the logging around snapshot to debugf | 19 August 2015, 00:18:08 UTC |
d01b6cd | Xiang Li | 18 August 2015, 23:49:43 UTC | Merge pull request #3326 from elimisteve/master client: fixed typo in WatcherOptions docs | 18 August 2015, 23:49:43 UTC |
9528271 | Steve Phillips | 18 August 2015, 23:42:52 UTC | client: fixed typo in WatcherOptions docs specifices -> specifies | 18 August 2015, 23:43:09 UTC |
b3d2a62 | Xiang Li | 18 August 2015, 23:35:13 UTC | Merge pull request #3325 from elimisteve/master client: spelling error in docs (occured -> occurred) | 18 August 2015, 23:35:13 UTC |
69fc796 | Steve Phillips | 18 August 2015, 23:26:52 UTC | client: spelling error in docs (occured -> occurred) | 18 August 2015, 23:26:52 UTC |
50c1db3 | Xiang Li | 18 August 2015, 22:43:49 UTC | raft: downgrade the logging around snapshot to debugf Snapshot related logging is spamming when leader trying to sync a failed peer. | 18 August 2015, 22:43:53 UTC |
7082d3a | Yicheng Qin | 18 August 2015, 05:27:26 UTC | docs: document how to set heartbeat interval and election timeout It gives more details about how to set heartbeat interval and election timeout correctly based on RTT. | 18 August 2015, 20:54:44 UTC |
28cec11 | Xiang Li | 18 August 2015, 19:38:51 UTC | Merge pull request #3322 from philips/use-proxy-as-default-endpoint Procfile: use proxy as default | 18 August 2015, 19:38:51 UTC |
087061e | Yicheng Qin | 18 August 2015, 19:06:48 UTC | Merge pull request #3303 from yichengq/auth-path use canonical path for auth | 18 August 2015, 19:06:48 UTC |
4778d78 | Yicheng Qin | 18 August 2015, 18:48:22 UTC | pkg/pathutil: change copyright for path.go The file only contains the function that is borrowed from std http lib, so we use their copyright. | 18 August 2015, 18:48:22 UTC |
9106675 | Brandon Philips | 18 August 2015, 16:36:57 UTC | Procfile: use proxy as default I think it makes sense to make the proxy listen on the default port so we can give the proxy more testing by default. Also, this should make it easy to kill a single etcd member and test that etcdctl still works, etc. However, I have hit a bug: the proxy takes several seconds | 18 August 2015, 16:42:13 UTC |
fab3fea | Yicheng Qin | 18 August 2015, 15:13:17 UTC | etcdctl/role: reject non-canonical permission path Non-canonical permission path is useless because the path received by auth is always canonical, which is due to our ServeMux always redirects request to canonical path(). This helps users to detect path permission setting error early. Ref: http://godoc.org/net/http#ServeMux | 18 August 2015, 15:59:53 UTC |
b5ec7f5 | Yicheng Qin | 14 August 2015, 23:14:23 UTC | client: use canonical url path in request The main change is that it keeps the trailing slash. This helps auth feature to judge path permission accurately. | 18 August 2015, 15:59:48 UTC |
927d5f3 | Yicheng Qin | 18 August 2015, 06:36:33 UTC | Merge pull request #3301 from yichengq/ca-file etcdmain: update -ca-file description | 18 August 2015, 06:36:33 UTC |
c0747a7 | Yicheng Qin | 14 August 2015, 20:58:22 UTC | etcdmain: update -ca-file description so people could deprecate old flags and use new flags much easier. | 18 August 2015, 05:36:04 UTC |
bcb4d5d | Yicheng Qin | 18 August 2015, 00:00:24 UTC | Merge pull request #3311 from yichengq/request-timeout extend hardcoded timeout for globally-deployed etcd cluster | 18 August 2015, 00:00:24 UTC |
dfc6b44 | Xiang Li | 17 August 2015, 23:54:12 UTC | Merge pull request #3315 from xiang90/key_err etcdhttp:write etcderror for all errors in keyhandler | 17 August 2015, 23:54:12 UTC |
ffae601 | Yicheng Qin | 17 August 2015, 21:10:05 UTC | etcdmain: calculate dial timeout for peer transport This helps peer communication in globally-deployed cluster. | 17 August 2015, 23:52:53 UTC |
1375ef8 | Yicheng Qin | 17 August 2015, 23:05:14 UTC | etcdserver: remove getVersion timeout The request can still time out because we have set dial timeout and read/write timeout. It increases timeout expectation from 1s to 5s, but it makes it workable in globally-deployer cluster. | 17 August 2015, 23:50:40 UTC |
c7fbc01 | Xiang Li | 17 August 2015, 23:04:00 UTC | Merge pull request #3314 from sebschrader/proxy-loop Warn about proxy loops with incorrect advertise-client-urls | 17 August 2015, 23:04:00 UTC |