I have an ElasticSearch 6.3 cluster running on 3 Azure nodes. Each node is 4 vcpus, 14 GB RAM. My snapshot repository is stored on an Azure fileshare. I have 36 indexes, 14Gb of data.
The snapshot repository is empty, but the snapshots are very slow (Even the initial one). When I check the performance of the cluster in Kibana, it shows typically 1% cpu and only 1Gb RAM used out of a possible 8Gb on each node. I do have compress set on the snapshot repository, but when I test tarring up all the indexes with a compress and then copying them onto the same fileshare, it takes a few minutes. The snapshot takes about an hour
Any experience / suggestions gratefully received