When storage is partitioned, refuse to serve requests unless web and databases agree on partitioning

Summary:
Ref T11044. One popular tool in a modern operations environment is Puppet. The primary purpose of this tool is to randomly revert hosts to older or different configurations.

Introducing an element of chaotic unpredictability into operations trains staff to be on high alert at all times, rather than lulled into complacency by predictability or consistency.

When Puppet reverts a Phabricator host's configuration to an older version, we might start writing data to a lot of crazy places where it shouldn't go. This will create a big sticky mess that is virtually impossible to undo, mostly because we'll get two files with ID 123 or two tasks with ID 456 or whatever else and good luck with that.

Instead, after changing the partition layout, require `bin/storage partition` to be run. This writes a copy of the config everywhere.

Then, when we start serving web requests, make sure every database has the exact same config. This will foil Puppet by refusing to run requests on hosts it has reverted.

Test Plan:
  - Changed partition configuration.
  - Ran Phabricator.
  - FOILED!
  - Ran `bin/storage partition` to sync config.
  - Things worked again.

Reviewers: chad

Reviewed By: chad

Maniphest Tasks: T11044

Differential Revision: https://secure.phabricator.com/D16910

resources/sql/autopatches/20161121.cluster.01.hoststate.sql

@@ -0,0 +1,5 @@
+CREATE TABLE {$NAMESPACE}_meta_data.hoststate (
+  stateKey VARCHAR(128) NOT NULL COLLATE {$COLLATE_TEXT},
+  stateValue LONGTEXT NOT NULL COLLATE {$COLLATE_TEXT},
+  PRIMARY KEY (stateKey)
+) ENGINE=InnoDB, COLLATE {$COLLATE_TEXT};

src/__phutil_library_map__.php

@@ -3795,6 +3795,7 @@ phutil_register_library_map(array(
     'PhabricatorStorageManagementDatabasesWorkflow' => 'infrastructure/storage/management/workflow/PhabricatorStorageManagementDatabasesWorkflow.php',
     'PhabricatorStorageManagementDestroyWorkflow' => 'infrastructure/storage/management/workflow/PhabricatorStorageManagementDestroyWorkflow.php',
     'PhabricatorStorageManagementDumpWorkflow' => 'infrastructure/storage/management/workflow/PhabricatorStorageManagementDumpWorkflow.php',
+    'PhabricatorStorageManagementPartitionWorkflow' => 'infrastructure/storage/management/workflow/PhabricatorStorageManagementPartitionWorkflow.php',
     'PhabricatorStorageManagementProbeWorkflow' => 'infrastructure/storage/management/workflow/PhabricatorStorageManagementProbeWorkflow.php',
     'PhabricatorStorageManagementQuickstartWorkflow' => 'infrastructure/storage/management/workflow/PhabricatorStorageManagementQuickstartWorkflow.php',
     'PhabricatorStorageManagementRenamespaceWorkflow' => 'infrastructure/storage/management/workflow/PhabricatorStorageManagementRenamespaceWorkflow.php',
@@ -8977,6 +8978,7 @@ phutil_register_library_map(array(
     'PhabricatorStorageManagementDatabasesWorkflow' => 'PhabricatorStorageManagementWorkflow',
     'PhabricatorStorageManagementDestroyWorkflow' => 'PhabricatorStorageManagementWorkflow',
     'PhabricatorStorageManagementDumpWorkflow' => 'PhabricatorStorageManagementWorkflow',
+    'PhabricatorStorageManagementPartitionWorkflow' => 'PhabricatorStorageManagementWorkflow',
     'PhabricatorStorageManagementProbeWorkflow' => 'PhabricatorStorageManagementWorkflow',
     'PhabricatorStorageManagementQuickstartWorkflow' => 'PhabricatorStorageManagementWorkflow',
     'PhabricatorStorageManagementRenamespaceWorkflow' => 'PhabricatorStorageManagementWorkflow',

src/applications/config/check/PhabricatorDatabaseSetupCheck.php

@@ -205,6 +205,38 @@ final class PhabricatorDatabaseSetupCheck extends PhabricatorSetupCheck {
         break;
     }
 
-
+    // If we have more than one master, we require that the cluster database
+    // configuration written to each database node is exactly the same as the
+    // one we are running with.
+    $masters = PhabricatorDatabaseRef::getAllMasterDatabaseRefs();
+    if (count($masters) > 1) {
+      $state_actual = queryfx_one(
+        $conn_meta,
+        'SELECT stateValue FROM %T WHERE stateKey = %s',
+        PhabricatorStorageManagementAPI::TABLE_HOSTSTATE,
+        'cluster.databases');
+      if ($state_actual) {
+        $state_actual = $state_actual['stateValue'];
       }
+
+      $state_expect = $ref->getPartitionStateForCommit();
+
+      if ($state_expect !== $state_actual) {
+        $message = pht(
+          'Database host "%s" has a configured cluster state which disagrees '.
+          'with the state on this host ("%s"). Run `bin/storage partition` '.
+          'to commit local state to the cluster. This host may have started '.
+          'with an out-of-date configuration.',
+          $ref->getRefKey(),
+          php_uname('n'));
+
+        $this->newIssue('db.state.desync')
+          ->setName(pht('Cluster Configuration Out of Sync'))
+          ->setMessage($message)
+          ->setIsFatal(true);
+        return true;
+      }
+    }
+  }
+
 }

src/docs/user/cluster/cluster_partitioning.diviner

@@ -123,6 +123,21 @@ Not all of the database partition names are the same as the application
 names. You can get a list of databases with `bin/storage databases` to identify
 the correct database names.
 
+After you have configured partitioning, it needs to be committed to the
+databases. This writes a copy of the configuration to tables on the databases,
+preventing errors if a webserver accidentally starts with an old or invalid
+configuration.
+
+To commit the configuration, run this command:
+
+```
+phabricator/ $ ./bin/storage partition
+```
+
+Run this command after making any partition or clustering changes. Webservers
+will not serve traffic if their configuration and the database configuration
+differ.
+
 
 Launching a new Partition
 =========================
@@ -135,6 +150,7 @@ To add a new partition, follow these steps:
     are partitioning, you will need to configure your existing master as the
     new "default". This will let Phabricator interact with it, but won't send
     any traffic to it yet.
+  - Run `bin/storage partition`.
   - Run `bin/storage upgrade` to initialize the schemata on the new hosts.
   - Stop writes to the applications you want to move by putting Phabricator
     in read-only mode, or shutting down the webserver and daemons, or telling
@@ -143,6 +159,7 @@ To add a new partition, follow these steps:
   - Load the data into the application databases on the new master.
   - Reconfigure the "partition" setup so that Phabricator knows the databases
     have moved.
+  - Run `bin/storage partition`.
   - While still in read-only mode, check that all the data appears to be
     intact.
   - Resume writes.

src/infrastructure/cluster/PhabricatorDatabaseRef.php

@@ -180,6 +180,17 @@ final class PhabricatorDatabaseRef
     return $this->applicationMap;
   }
 
+  public function getPartitionStateForCommit() {
+    $state = PhabricatorEnv::getEnvConfig('cluster.databases');
+    foreach ($state as $key => $value) {
+      // Don't store passwords, since we don't care if they differ and
+      // users may find it surprising.
+      unset($state[$key]['pass']);
+    }
+
+    return phutil_json_encode($state);
+  }
+
   public function setMasterRef(PhabricatorDatabaseRef $master_ref) {
     $this->masterRef = $master_ref;
     return $this;
@@ -498,9 +509,6 @@ final class PhabricatorDatabaseRef
 
     $masters = array();
     foreach ($refs as $ref) {
-      if ($ref->getDisabled()) {
-        continue;
-      }
       if ($ref->getIsMaster()) {
         $masters[] = $ref;
       }

src/infrastructure/storage/management/PhabricatorStorageManagementAPI.php

@@ -19,6 +19,7 @@ final class PhabricatorStorageManagementAPI extends Phobject {
   const COLLATE_FULLTEXT = 'COLLATE_FULLTEXT';
 
   const TABLE_STATUS = 'patch_status';
+  const TABLE_HOSTSTATE = 'hoststate';
 
   public function setDisableUTF8MB4($disable_utf8_mb4) {
     $this->disableUTF8MB4 = $disable_utf8_mb4;

src/infrastructure/storage/management/workflow/PhabricatorStorageManagementPartitionWorkflow.php

@@ -0,0 +1,44 @@
+<?php
+
+final class PhabricatorStorageManagementPartitionWorkflow
+  extends PhabricatorStorageManagementWorkflow {
+
+  protected function didConstruct() {
+    $this
+      ->setName('partition')
+      ->setExamples('**partition** [__options__]')
+      ->setSynopsis(pht('Commit partition configuration to databases.'))
+      ->setArguments(array());
+  }
+
+  public function didExecute(PhutilArgumentParser $args) {
+    echo tsprintf(
+      "%s\n",
+      pht('Committing configured partition map to databases...'));
+
+    foreach ($this->getMasterAPIs() as $api) {
+      $ref = $api->getRef();
+      $conn = $ref->newManagementConnection();
+
+      $state = $ref->getPartitionStateForCommit();
+
+      queryfx(
+        $conn,
+        'INSERT INTO %T.%T (stateKey, stateValue) VALUES (%s, %s)
+          ON DUPLICATE KEY UPDATE stateValue = VALUES(stateValue)',
+        $api->getDatabaseName('meta_data'),
+        PhabricatorStorageManagementAPI::TABLE_HOSTSTATE,
+        'cluster.databases',
+        $state);
+
+      echo tsprintf(
+        "%s\n",
+        pht(
+          'Wrote configuration on database host "%s".',
+          $ref->getRefKey()));
+    }
+
+    return 0;
+  }
+
+}

src/infrastructure/storage/schema/PhabricatorStorageSchemaSpec.php

@@ -18,6 +18,20 @@ final class PhabricatorStorageSchemaSpec
           'unique' => true,
         ),
       ));
+
+    $this->buildRawSchema(
+      'meta_data',
+      PhabricatorStorageManagementAPI::TABLE_HOSTSTATE,
+      array(
+        'stateKey' => 'text128',
+        'stateValue' => 'text',
+      ),
+      array(
+        'PRIMARY' => array(
+          'columns' => array('stateKey'),
+          'unique' => true,
+        ),
+      ));
   }
 
 }