How i to analysis the es cluster?

I find my es cluster sometimes index docs(use bulk api) very slow.
At that time,I use cat thread_pool api find are all less 2(bulk.size are all 32).And i find my server's CPU's usage are also very little.And the disk's write/per second just nearly 80M/s (3 800G+ SSD disk use raid0).And there are no query at that time.
How i to find out why es index slow?


How slow?
How large are your bulk requests?

@warkolm,Thank you!

Normally,my es cluster(6 nodes) can index 10w/s,but some times it just 1-2w/s.
Bulk batchSize 1500,average pre doc's size is 1K.

What is w/s?


I get the Indexing Rate from marvel,just like '100000/s'.

Right, but I don't know what the w has to do with it.
Is it 10 per second?

@warkolm 100000 per second. . Sorry ,we say ten thousand as w..