Datastreams losing indices

Elastic Search
1

Hi,

I'm facing some strange things with my ELK stack.
Since 2/3 days, some datastreams have lost some indices, storage size went to 0 docs and 0b ! Those indices are random in the timeline.
It also seems to not write anymore in new indices...

Capture d’écran 2024-10-25 101044
Capture d’écran 2024-10-25 1010441566×653 79.6 KB

Any idea what could go wrong ?

P.S.: At the moment, the elastic disk storage is full at 90% (100Go left) on the hot node, and 85% (73Go left) on the warm node.

2

I would start from looking into logs for Elasticsearch - it's often pretty verbose, especially in DEBUG log level.

Can you look at the logs and maybe grep by the index names to see what Elasticsearch reported?

3

After investigating, it seems that when you reach the 90% watermark, elastic try to free space by moving some indices to another 'same' node (hot/warm/cold), and stop writing/indexing new entries.
But in my case, I only have 1 hot and 1 warm node, without replicas.

I managed to delete some old indices to gain free space, and elastic start rebalancing the indices.
To solve this with a more long term solution, I guess I have to increase the size of my data disks.

Is it possible to change those watermarks to 95% for example, because I loose near 100Go of disk space on the hot node with this one ?

4

You can indeed update watermarks following the docs: Fix watermark errors | Elasticsearch Guide [8.15] | Elastic

This should be the setting you're looking for: Cluster-level shard allocation and routing settings | Elasticsearch Guide [8.15] | Elastic

cluster.routing.allocation.disk.watermark.high logo cloud
(Dynamic) Controls the high watermark. It defaults to 90%, meaning that Elasticsearch will attempt to relocate shards away from a node whose disk usage is above 90%. It can alternatively be set to a ratio value, e.g., 0.9. It can also be set to an absolute byte value (similarly to the low watermark) to relocate shards away from a node if it has less than the specified amount of free space. This setting affects the allocation of all shards, whether previously allocated or not.

PUT _cluster/settings
{
  "persistent": {
    "cluster.routing.allocation.disk.watermark.high": "95%"
  }
}

However, it's still highly recommended to increase number of nodes or increase the storage size - at some point you'll reach this watermark and there will be no way out of it.