rolling_upgrade.asciidoc 4.7 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159
  1. [[rolling-upgrades]]
  2. == Rolling upgrades
  3. A rolling upgrade allows an Elasticsearch cluster to be upgraded one node at
  4. a time so upgrading does not interrupt service. Running multiple versions of
  5. Elasticsearch in the same cluster beyond the duration of an upgrade is
  6. not supported, as shards cannot be replicated from upgraded nodes to nodes
  7. running the older version.
  8. Rolling upgrades can be performed between minor versions. Elasticsearch
  9. 6.x supports rolling upgrades from *Elasticsearch 5.6*.
  10. Upgrading from earlier 5.x versions requires a <<restart-upgrade,
  11. full cluster restart>>. You must <<reindex-upgrade,reindex to upgrade>> from
  12. versions prior to 5.x.
  13. To perform a rolling upgrade:
  14. . *Disable shard allocation*.
  15. +
  16. --
  17. include::disable-shard-alloc.asciidoc[]
  18. --
  19. . *Stop non-essential indexing and perform a synced flush.* (Optional)
  20. +
  21. --
  22. While you can continue indexing during the upgrade, shard recovery
  23. is much faster if you temporarily stop non-essential indexing and perform a
  24. <<indices-synced-flush, synced-flush>>.
  25. include::synced-flush.asciidoc[]
  26. --
  27. . [[upgrade-node]] *Shut down a single node*.
  28. +
  29. --
  30. include::shut-down-node.asciidoc[]
  31. --
  32. . *Upgrade the node you shut down.*
  33. +
  34. --
  35. include::upgrade-node.asciidoc[]
  36. include::set-paths-tip.asciidoc[]
  37. --
  38. . *Upgrade any plugins.*
  39. +
  40. Use the `elasticsearch-plugin` script to install the upgraded version of each
  41. installed Elasticsearch plugin. All plugins must be upgraded when you upgrade
  42. a node.
  43. . *Start the upgraded node.*
  44. +
  45. --
  46. Start the newly-upgraded node and confirm that it joins the cluster by checking
  47. the log file or by submitting a `_cat/nodes` request:
  48. [source,sh]
  49. --------------------------------------------------
  50. GET _cat/nodes
  51. --------------------------------------------------
  52. // CONSOLE
  53. --
  54. . *Reenable shard allocation.*
  55. +
  56. --
  57. Once the node has joined the cluster, reenable shard allocation to start using
  58. the node:
  59. [source,js]
  60. --------------------------------------------------
  61. PUT _cluster/settings
  62. {
  63. "transient": {
  64. "cluster.routing.allocation.enable": "all"
  65. }
  66. }
  67. --------------------------------------------------
  68. // CONSOLE
  69. --
  70. . *Wait for the node to recover.*
  71. +
  72. --
  73. Before upgrading the next node, wait for the cluster to finish shard allocation.
  74. You can check progress by submitting a <<cat-health,`_cat/health`>> request:
  75. [source,sh]
  76. --------------------------------------------------
  77. GET _cat/health
  78. --------------------------------------------------
  79. // CONSOLE
  80. Wait for the `status` column to switch from `yellow` to `green`. Once the
  81. node is `green`, all primary and replica shards have been allocated.
  82. [IMPORTANT]
  83. ====================================================
  84. During a rolling upgrade, primary shards assigned to a node running the new
  85. version cannot have their replicas assigned to a node with the old
  86. version. The new version might have a different data format that is
  87. not understood by the old version.
  88. If it is not possible to assign the replica shards to another node
  89. (there is only one upgraded node in the cluster), the replica
  90. shards remain unassigned and status stays `yellow`.
  91. In this case, you can proceed once there are no initializing or relocating shards
  92. (check the `init` and `relo` columns).
  93. As soon as another node is upgraded, the replicas can be assigned and the
  94. status will change to `green`.
  95. ====================================================
  96. Shards that were not <<indices-synced-flush,sync-flushed>> might take longer to
  97. recover. You can monitor the recovery status of individual shards by
  98. submitting a <<cat-recovery,`_cat/recovery`>> request:
  99. [source,sh]
  100. --------------------------------------------------
  101. GET _cat/recovery
  102. --------------------------------------------------
  103. // CONSOLE
  104. If you stopped indexing, it is safe to resume indexing as soon as
  105. recovery completes.
  106. --
  107. . *Repeat*
  108. +
  109. --
  110. When the node has recovered and the cluster is stable, repeat these steps
  111. for each node that needs to be updated.
  112. --
  113. [IMPORTANT]
  114. ====================================================
  115. During a rolling upgrade, the cluster continues to operate normally. However,
  116. any new functionality is disabled or operates in a backward compatible mode
  117. until all nodes in the cluster are upgraded. New functionality
  118. becomes operational once the upgrade is complete and all nodes are running the
  119. new version. Once that has happened, there's no way to return to operating
  120. in a backward compatible mode. Nodes running the previous major version will
  121. not be allowed to join the fully-updated cluster.
  122. In the unlikely case of a network malfunction during the upgrade process that
  123. isolates all remaining old nodes from the cluster, you must take the
  124. old nodes offline and upgrade them to enable them to join the cluster.
  125. ====================================================