forked from docs/doc-exports
zhengxiu
93d856d5c5
Reviewed-by: Pruthi, Vineet <vineet.pruthi@t-systems.com> Co-authored-by: zhengxiu <zhengxiu@huawei.com> Co-committed-by: zhengxiu <zhengxiu@huawei.com>
16 lines
2.7 KiB
HTML
16 lines
2.7 KiB
HTML
<a name="EN-US_TOPIC_0000001933159150"></a><a name="EN-US_TOPIC_0000001933159150"></a>
|
|
|
|
<h1 class="topictitle1">Will CSS Cluster Services Be Affected If the Disk Usage of a Single Node Gets Too High?</h1>
|
|
<div id="body0000001591144897"><div class="section" id="EN-US_TOPIC_0000001933159150__section0221183314557"><h4 class="sectiontitle">Symptom</h4><p id="EN-US_TOPIC_0000001933159150__p16439123714555">According to the cluster monitoring information, the disk usage of some individual nodes of an Elasticsearch cluster exceeds 80%. Does it impact cluster performance?</p>
|
|
</div>
|
|
<div class="section" id="EN-US_TOPIC_0000001933159150__section75501729195813"><h4 class="sectiontitle">Impact on Services</h4><ul id="EN-US_TOPIC_0000001933159150__ul95121359175215"><li id="EN-US_TOPIC_0000001933159150__li35125598522">When the disk usage of a node exceeds 85%, disk space cannot be allocated to new replicas, but can still be allocated to new primary shards. This ensures service continuity, but it impacts the high availability of the Elasticsearch cluster.</li><li id="EN-US_TOPIC_0000001933159150__li65129591521">When the disk usage of a node exceeds 90%, a shard migration is automatically triggered to reallocate shards on this node to other data nodes with lower disk usage. During this process, disk space cannot be allocated to new shards. Shard migration and reallocation may increase query delay or temporarily interrupt services, impacting service continuity.</li><li id="EN-US_TOPIC_0000001933159150__li440313481855">When the disk usage of a node exceeds 95%, the <span class="parmname" id="EN-US_TOPIC_0000001933159150__parmname176121424177"><b>read_only_allow_delete</b></span> attribute is enabled in its indexes. In this case, indexes on this node can only be read or deleted but data cannot be written in.</li></ul>
|
|
<p id="EN-US_TOPIC_0000001933159150__p451511614155">If per-node disk usage is too high, you can add more nodes or expand the capacity of existing nodes. For details, see <a href="css_01_0151.html">Scaling Out/Up an Elasticsearch Cluster</a>. After new nodes are added, index shards automatically re-allocate to balance the load. You can launch Cerebro to check the shard allocation process. You can also modify <span class="parmname" id="EN-US_TOPIC_0000001933159150__parmname51458329720"><b>indices.recovery.max_bytes_per_sec</b></span> and <span class="parmname" id="EN-US_TOPIC_0000001933159150__parmname1982313618713"><b>cluster.routing.allocation.cluster_concurrent_rebalance</b></span> to speed up this process.</p>
|
|
</div>
|
|
</div>
|
|
<div>
|
|
<div class="familylinks">
|
|
<div class="parentlink"><strong>Parent topic:</strong> <a href="css_02_0148.html">CSS Cluster Monitoring and O&M</a></div>
|
|
</div>
|
|
</div>
|
|
|