In case of URLs that are synonyms, the current lexicographic sorting
and compare of the URLs fails with frustrating errors. Make sure to do
a full comparison between every set of PeerURLs before failing.

Fixes etcd-io#11013

Use golang.org/x/sys/unix for F_OFD_* constants.

This fixes the issue that F_OFD_GETLK was defined incorrectly,
resulting in bugs such as moby/moby#31182

Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>

[3.4 backport] pkg/fileutil: fix F_OFD_ constants

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

vendor: bump gorilla/websocket

This fixes etcd being unable to send any message longer than 64 KB as
a notification over the websocket. This was because the older version
of grpc-websocket-proxy was used and WithMaxRespBodyBufferSize option
wasn't set.

etcdserver: Fix 64 KB websocket notification message limit

…de health check in debug level

ref. etcd-io#12677
ref. etcd-io@0b9cfa8

[Backport-3.4] etcdserver/api/etcdhttp: log successful etcd server side health check in debug level

Signed-off-by: Gyuho Lee <leegyuho@amazon.com>

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

Manual cherry pick of etcd-io#12448 on release 3.4

There are situations where we don't wish to fsync but we do want to
write the data.

Typically this occurs in clusters where fsync latency (often the
result of firmware) transiently spikes.  For Kubernetes clusters this
causes (many) elections which have knock-on effects such that the API
server will transiently fail causing other components fail in turn.

By writing the data (buffered and asynchronously flushed, so in most
situations the write is fast) and avoiding the fsync we no longer
trigger this situation and opportunistically write out the data.

Anecdotally:
  Because the fsync is missing there is the argument that certain
  types of failure events will cause data corruption or loss, in
  testing this wasn't seen.  If this was to occur the expectation is
  the member can be readded to a cluster or worst-case restored from a
  robust persisted snapshot.

  The etcd members are deployed across isolated racks with different
  power feeds.  An instantaneous failure of all of them simultaneously
  is unlikely.

  Testing was usually of the form:
   * create (Kubernetes) etcd write-churn by creating replicasets of
     some 1000s of pods
   * break/fail the leader

  Failure testing included:
   * hard node power-off events
   * disk removal
   * orderly reboots/shutdown

  In all cases when the node recovered it was able to rejoin the
  cluster and synchronize.

When using --unsafe-no-fsync still write out the data

The integration jobs fail with timeouts slightly over 3s, increase
this marginally so false failures are less prevalent.

integration: relax leader timeout from 3s to 4s

…tion

etcdserver: Fix PeerURL validation

Manual cherry-pick of 9571325 for
release-3.4.

etcdserver: fix incorrect metrics generated when clients cancel watches

As go 1.12.2 is what is tested in CI as well as recommended to be built
with 1.12.2 we should also pin to this in the go directive version.

[release-3.4]: Pin go version in go.mod to 1.12

Currently in CI the tests are only run with go v1.12, this adds also go
v1.15.11.

Excludes certain variants for v1.15.

This patch is needed due to go 1.15 erroring on:

"Setctty set but Ctty not valid in child".

In go1.13, the TLS13 is enablled by default, and as per go1.13 release notes :
TLS 1.3 cipher suites are not configurable. All supported cipher suites are safe,
and if PreferServerCipherSuites is set in Config the preference order is based
on the available hardware.

Fixing the test case for go1.13 by limiting the TLS version to TLS12

Currently with race it fails, we can enable this at a later point.

In fact this commit rewrites the functionality to use upstream list of
ciphers instead of checking whether the lists are in sync using ast
analysis.

Cherry-pick of 60e4428 from master branch does not work due to
missing `tls.CipherSuites()` function. We work around by using go build
tags for both the building and tests.

…egration/...'

grpc proxy opens additional 2 watching channels. The metric is shared
between etcd-server & grpc_proxy, so all assertions on number of open
watch channels need to take in consideration the additional "2"
channels.

[release-3.4]: .travis.yml: Test with go v1.15.11

This version was already used to build the release v3.4.15.

.travis,Makefile,functional: Bump go 1.12 version to v1.12.17

…uests

…k-12871

Backport-3.4 etcdserver/util.go: reduce memory when logging range requests

backport PR etcd-io#12894 to release-3.4

[Backport-3.4] etcdserver/mvcc: update trace.Step condition

…12880

Backport-3.4 exclude alarms from health check conditionally

Signed-off-by: Gyuho Lee <leegyuho@amazon.com>

…r when its cert has non empty CN

CARRY:

This patch is required to allow metrics cert to auth with CN in cert.

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

CARRY:

During the release of 3.4.9 gyuho bumped go.mod to 1.14. Since the release is expected to use 1.12 we will not keep the commit.

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

…tionally"

This reverts commit dbde4f2.

…ul etcd server side health check in debug level

This reverts commit f27ef4d.

This reverts commit cfc08e5.

This reverts commit b60e76b.

This reverts commit 0e2dafd.

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

… is a distributed"

Correcting a typo from 2f10964 (version: openshift-v4.0, 2018-11-29).

…operly copied from build.

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

…se-3.3

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

make build performs a sanity test on the binary image which causes problems for unsupport arch. Because we run full CI tests against the image this check is not nessisary and will allow images to be build regardless of arch.

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

…ember but memberDir exists, go ahead and start.

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

…dling when we dont scale member

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

…nsistent with ART

Reconciling with https://github.com/openshift/ocp-build-data/tree/990044f295fb1d5e238823902962dbcfa1c041c9/images/ose-etcd.yml

…nsistent with ART

Reconciling with https://github.com/openshift/ocp-build-data/tree/ac81dd4ff0bd57c4e75058d25b40615b92948259/images/ose-etcd.yml

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

…nsistent with ART

Reconciling with https://github.com/openshift/ocp-build-data/tree/5a1293dd0f380abf50c12d65c36655486d7745d0/images/ose-etcd.yml

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

To make it easier to root-cause when /health check fails.
For example, we are using load balancer to health check
each etcd instance, and when one etcd node gets terminated,
it's hard to tell whether etcd "server" was really failing
or client (or load balancer") failed to reach the etcd cluster
which is also failure in load balancer health check.

Signed-off-by: Gyuho Lee <leegyuho@amazon.com>

…r side health check in

debug level

…s not part of member list and dataDir exists

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>

Reconciling with https://github.com/openshift/ocp-build-data/tree/691e628254f318ce56efda5edc7448ec743c37b8/images/ose-etcd.yml

Signed-off-by: Sam Batschelet <sbatsche@redhat.com>