VTOrc is the automated fault detection and repair tool of Vitess. It started off as a fork of the Orchestrator, which was then custom-fitted to the Vitess use-case running as a Vitess component. It has reached general availablity with this release of Vitess. An overview of the architecture of VTOrc can be found on this page.

Setting up VTOrc lets you avoid performing the InitShardPrimary step. It automatically detects that the new shard doesn't have a primary and elects one for you.

Configuration Refactor and New Flags #

Since VTOrc was forked from Orchestrator, it inherited a lot of configurations that don't make sense for the Vitess use-case. All of such configurations have been removed.

For all the configurations that are kept, flags have been added for them and the flags are the desired way to pass these configurations going forward. The config file will be deprecated and removed in upcoming releases. The following is a list of all the configurations that are kept and the associated flags added.

Configurations KeptFlags IntroducedFlag Usage
SQLite3DataFile--sqlite-data-fileSQLite Datafile to use as VTOrc's database
InstancePollSeconds--instance-poll-timeTimer duration on which VTOrc refreshes MySQL information
SnapshotTopologiesIntervalHours--snapshot-topology-intervalTimer duration on which VTOrc takes a snapshot of the current MySQL information it has in the database. Should be in multiple of hours
ReasonableReplicationLagSeconds--reasonable-replication-lagMaximum replication lag on replicas which is deemed to be acceptable
AuditLogFile--audit-file-locationFile location where the audit logs are to be stored
AuditToSyslog--audit-to-backendWhether to store the audit log in the VTOrc database
AuditToBackendDB--audit-to-syslogWhether to store the audit log in the syslog
AuditPurgeDays--audit-purge-durationDuration for which audit logs are held before being purged. Should be in multiples of days
RecoveryPeriodBlockSeconds--recovery-period-block-durationDuration for which a new recovery is blocked on an instance after running a recovery
PreventCrossDataCenterPrimaryFailover--prevent-cross-cell-failoverPrevent VTOrc from promoting a primary in a different cell than the current primary in case of a failover
LockShardTimeoutSeconds--lock-shard-timeoutDuration for which a shard lock is held when running a recovery
WaitReplicasTimeoutSeconds--wait-replicas-timeoutDuration for which to wait for replica's to respond when issuing RPCs
TopoInformationRefreshSeconds--topo-information-refresh-durationTimer duration on which VTOrc refreshes the keyspace and vttablet records from the topology server
RecoveryPollSeconds--recovery-poll-durationTimer duration on which VTOrc polls its database to run a recovery

For a full list of supported flags, please look at VTOrc reference page.

Apart from configurations, some flags from VTOrc have also been removed -

  • sibling
  • destination
  • discovery
  • skip-unresolve
  • skip-unresolve-check
  • noop
  • binlog
  • statement
  • grab-election
  • promotion-rule
  • skip-continuous-registration
  • enable-database-update
  • ignore-raft-setup
  • tag

Old UI Removal and Replacement #

The old UI that VTOrc inherited from Orchestrator has been removed. A debug UI, more consistent with the other Vitess binaries has been created. In order to use the new UI, --port flag has to be provided.

Along with the UI, the old APIs have also been deprecated. However, some of them have been ported over to the new UI -

Old APINew APIAdditional notes
/api/problems/api/problemsThis API lists all the instances that have any problems in them. The problems range from replication not running to errant GTIDs. The new API also supports filtering using the keyspace and shard name
/api/disable-global-recoveries/api/disable-global-recoveriesThis API disables the global recoveries in VTOrc. This makes it so that VTOrc doesn't repair any failures it detects.
/api/enable-global-recoveries/api/enable-global-recoveriesThis API enables the global recoveries in VTOrc.
/api/health/debug/healthThis API outputs the health of the VTOrc process.
/api/replication-analysis/api/replication-analysisThis API shows the replication analysis of VTOrc. Output is in JSON format.

Apart from these APIs, we also now have /debug/status, /debug/vars and /debug/liveness available in the new UI.

For more information about the UI, API and metrics that VTOrc exports, please consult this page.

In order to change the primary tablet of a running cluster, instead of drag and drop from the old UI or using the graceful-primary-takeover API, please use VTAdmin or vtctldclient to execute PlannedReparentShard.

Example invocation of VTOrc #

You can bring VTOrc using the following invocation:

vtorc --topo_implementation etcd2 \
  --topo_global_server_address "localhost:2379" \
  --topo_global_root /vitess/global \
  --port 15000 \
  --log_dir=${VTDATAROOT}/tmp \
  --recovery-period-block-duration "10m" \
  --instance-poll-time "1s" \
  --topo-information-refresh-duration "30s" \

You can optionally add a clusters_to_watch flag that contains a comma separated list of keyspaces or keyspace/shard values. If specified, VTOrc will manage only those clusters.

Durability Policies #

All the failovers that VTOrc performs will be honoring the durability policies. Please be careful in setting the desired durability policies for your keyspace because this will affect what situations VTOrc can recover from and what situations will require manual intervention.

Example Upgrade From v14 #

If you are running VTOrc with the flags --ignore-raft-setup --clusters_to_watch="ks/0" --config="path/to/config" and the following configuration

  "Debug": true,
  "ListenAddress": ":6922",
  "MySQLTopologyUser": "orc_client_user",
  "MySQLTopologyPassword": "orc_client_user_password",
  "MySQLReplicaUser": "vt_repl",
  "MySQLReplicaPassword": "",
  "RecoveryPeriodBlockSeconds": 1,
  "InstancePollSeconds": 1,
  "PreventCrossDataCenterPrimaryFailover": true

First drop the flag --ignore-raft-setup while on the previous release, since it is no longer available in this release. So, you'll be running VTOrc with --clusters_to_watch="ks/0" --config="path/to/config" and the same configuration listed above.

Now you can upgrade your VTOrc version continuing to use the same flags and configurations, and it will continue to work just the same. If you wish to use the new UI and APIs, then you can add the --port flag as well.

After upgrading, you can drop the configuration entirely and only use the new flags like --clusters_to_watch="ks/0" --recovery-period-block-duration=1s --instance-poll-time=1s --prevent-cross-cell-failover. This is the desired state because the support for the configuration file will be removed in upcoming releases.

Running VTOrc using the Vitess Operator #

To find information about deploying VTOrc using Vitess Operator please take a look at this page.