ELK on Windows: Very Poor Performance

[gokhankocak]

We setup a cluster with 2 data nodes on Hyper-V VMs.
Daily index size is around 200 GB.
Performance is very poor, elasticsearch does not respond at all.
2 data nodes use 100% CPU and 32 GB of memory (each node).
We use the latest version, 5.3.0 with the latest Java version.

I also observed 400 Mbit/s network traffic between nodes.

We need some guidance on sizing.
Or should we setup Linux nodes instead of Windows?

[warkolm (Mark Walkom)]

How many indices, how many shards?
Do you have Monitoring enabled?

[gokhankocak]

Hi Mark,

Yes, we have monitoring enabled.
There are 288 primary shards and 283 replica shards and 68 indices.
Indices keep daily logs, so the latest one is most active.

[cid:image003.jpg@01D2B8ED.C8064620]

[cid:image004.jpg@01D2B8ED.C8064620]

[cid:image005.jpg@01D2B8ED.C8064620]

// Gökhan Koçak
// Uygulama Yönetimi Mimari Danışmanı