summaryrefslogtreecommitdiffstats
path: root/zookeeper-server/zookeeper-server-common
Commit message (Collapse)AuthorAgeFilesLines
* Generate server ZK TLS config using Vespa mTLS configBjørn Christian Seime2021-04-213-122/+67
| | | | Server ZK TLS config follows Vespa mTLS config and is no longer controlled by feature flag.
* Another attempt at configuring secure only client port on ZK serverBjørn Christian Seime2021-03-182-30/+23
| | | | | | ZK dynamic reconfiguration logic assumes that insecure client port exists. This commit introduces a new connection factory that overrides 'secure' flag from configure() and makes the insecure client port become secure.
* Use default metrics providerHarald Musum2021-03-182-3/+1
|
* Update testsHarald Musum2021-03-171-6/+0
|
* Set auth provider as system propertuy and use x509 as schemeHarald Musum2021-03-171-3/+2
| | | | Using config does not work, need to use system property
* Use custom x509 authentication provider for ZK serverBjørn Christian Seime2021-03-172-4/+12
| | | | | Default provider implementation from ZK does not work in conjunction with ssl context supplier (fails on missing trust manager).
* Remove unused parameterBjørn Christian Seime2021-03-151-3/+3
|
* Handle client port that is configured to TLS onlyBjørn Christian Seime2021-03-152-21/+35
| | | | | | The client port can no longer be distributed through the ZK dynamic reconfiguration as the protocol does not support SSL client port. The port must be configured through the static config section instead.
* Use SSLContext supplier for ZK server configurationBjørn Christian Seime2021-03-124-231/+90
| | | | | Remove use of keystore/truststore from disk. Only configure client secure port if TLS is enabled.
* Simplify and stabilise test (would fail if asleep for more than 1s)Jon Marius Venstad2021-03-121-16/+5
|
* Use correct List typeBjørn Christian Seime2021-03-101-1/+1
|
* Revert "Revert "Specify TLS configuration when enabling secure ZK client""Bjørn Christian Seime2021-03-103-35/+58
|
* Revert "Specify TLS configuration when enabling secure ZK client"Harald Musum2021-03-103-58/+35
|
* Specify TLS configuration when enabling secure ZK clientBjørn Christian Seime2021-03-103-35/+58
| | | | | Implement and use ssl context supplier class. Move helpers methods for determining enabled ciphers/protocols to supplier class.
* Revert "Periodically reconfigure config server ZooKeeper cluster"Martin Polden2021-03-013-81/+4
|
* Ignore config with no serversMartin Polden2021-02-262-0/+6
|
* Support reconfiguration without configMartin Polden2021-02-263-4/+75
|
* Disable TLSV1.3Bjørn Christian Seime2021-02-241-2/+2
|
* Flag service for orchestrated restart when ZK myid changesJon Marius Venstad2021-02-231-1/+1
|
* Fix testsJon Marius Venstad2021-02-191-18/+24
|
* Move ownership of QuorumPeer to RecoonfigurerJon Marius Venstad2021-02-193-12/+49
|
* Die after 1 minute if no ZK-connection when constructing CCJon Marius Venstad2021-02-191-2/+2
|
* ZK will now allow TLSv1.3 in addition to TLSv1.2Bjørn Christian Seime2021-02-181-2/+2
|
* Merge pull request #16557 from ↵Jon Marius Venstad2021-02-172-15/+26
|\ | | | | | | | | vespa-engine/jonmv/properly-detect-leaving-zk-servers Properly detect leaving ZK servers by comparing full spec
| * Properly detect leaving ZK servers by comparing full specJon Marius Venstad2021-02-172-15/+26
| |
* | Merge pull request #16555 from vespa-engine/hmusum/use-listJon Marius Venstad2021-02-171-2/+5
|\ \ | |/ |/| Use a list instead of a set, we want to keep order
| * Add distinct()Harald Musum2021-02-171-1/+4
| |
| * Use a list instead of a set, we want to keep orderHarald Musum2021-02-171-2/+2
| |
* | Print actual servers from config and reformatHarald Musum2021-02-171-4/+4
| |
* | Log active and new config before reconfiguringHarald Musum2021-02-171-1/+3
|/
* Shutdown and die if we are going down to 1 ZooKeeper serverHarald Musum2021-02-161-0/+8
| | | | Reconfiguration will not work, since the old servers will be shutdown
* Do logAndDie in reconfigure methodHarald Musum2021-02-111-8/+8
|
* Exit if reconmfiguration failsHarald Musum2021-02-111-4/+11
| | | | | | | | | When going from 2 to 1 zookeeper servers (after going from 1 to 2 nodes because a node is about to be retired), reconfiguration will fail if the second node is removed before reconfiguration has happened. Reconfiguration will never work in these cases, since 2 nodes is needed to have ZooKeepere quorum. Just exit if this happens, new config will be used in this case and ZooKeeper should work fine again.
* Always force shutdown of non-reconfigurable ZooKeeper serverMartin Polden2020-12-164-6/+32
| | | | | This basically reverts to the original behaviour, where the container is forced to exit when a new VespaZooKeeperServer component is setup.
* Use exponential backoff for ZooKeeper restartMartin Polden2020-12-154-29/+53
|
* Shut down ZooKeeper properlyMartin Polden2020-12-154-2/+10
|
* Never attempt to start again after executor has shut downMartin Polden2020-12-141-1/+1
|
* Log when triggering shutdownMartin Polden2020-12-141-2/+5
|
* Log exceptionHarald Musum2020-12-141-1/+1
|
* Log time left on reconfig failureMartin Polden2020-12-131-3/+7
|
* Adjust total timeout according to joining serversMartin Polden2020-12-131-3/+12
|
* Prevent tests from blocking until shutdown timeoutMartin Polden2020-12-133-53/+52
|
* Merge pull request #15791 from vespa-engine/mpolden/joining-serversMartin Polden2020-12-112-56/+71
|\ | | | | Make joining server an observer in initial config
| * Make joining server an observer in initial configMartin Polden2020-12-112-10/+20
| |
| * Clean up testMartin Polden2020-12-101-47/+52
| |
* | Remove duplicate log message now that we retry if starting failsHarald Musum2020-12-111-1/+0
| |
* | Retry start of zookeeper serverHarald Musum2020-12-101-6/+24
|/ | | | | An already running server may take som time to shut down, starting a new one might fail in that case so we need to retry
* 2 hacks to be able to use ZooKeeper server 3.6Harald Musum2020-12-093-2/+12
| | | | | | | | | | | Explicitly depend on io.dropwizard.metrics and org.xerial.snappy since they are used by metrics provider (even NullMetricsProvider) when bootstrapping ServerMetrics in zookeeper code and cannot be found when runnin in a container Do some class loadin trick to circumvent zookeeper itself loading a class with the threads current class loader, which does not work well in a container
* Log error when starting ZooKeeper server failsHarald Musum2020-12-091-1/+5
|
* Revert "Set reconfigEnabled based on config value"Harald Musum2020-12-092-3/+1
|