hi James,

it used to be the same server, zookeeper included in node1.

I even setup  a new zookeeper root to reset all configs and create a new "drill 
cluster" but querying was still slow.

I then created a third node in the same vnet, subnet, same vm size, but this 
time I didn't join it to the cluster.
So in embeded mode it queried the same sources, this time the planning step 
took ~3secs (which is the expected time i guess).

Given that I couldn't recover the first cluster I decided to create a separate 
vm for zookeeper and connected a new drill node there,  now the planner takes 
again ~3 secs even in distributed mode.

So all good on my side, but I'm still puzzled  as to what can cause the planner 
to take so long?
In the documentation it says that it could be the plugins so I disabled them 
all and that didn't change anything.

This info could be nice to have before attempting to go to prod, do you have 
any ideas?

thanks again for your reply!

Jorge
________________________________
From: James Turton <[email protected]>
Sent: Monday, January 17, 2022 10:22
To: [email protected] <[email protected]>; Jorge Alvarado 
<[email protected]>
Subject: Re: drill planning taking always ~30secs

Hi Jorge

When you get good results using embedded Drill are you running it on the
same machines where you have created your cluster?

James

On 2022/01/16 14:34, Jorge Alvarado wrote:
> hi Drill community,
>
> I was wondering if you can help again on my discovery of Drill.
> I had been successfully querying multiple source such as SQL Server, MySQl, 
> AWS S3, even SAP HANA.
> When I used the embedded run, it worked like a charm.
> Then I tried to configure a small cluster with 2 drillbits and 1 zookeeper 
> (same cloud region, same VNet, same subnet)
> Then I've noticed the queries are taking quite a long time (~30 secs)
> And looking further into the querying profile I see the most part is spent on 
> the query planning,
> which doesn't make much sense to me as the sources haven't changed, and the 
> queries are very simple (SELECT * FROM xxx)
> Is there something I can do to "optimize" the planner? I cannot find a good 
> setting to update.
>
> Thanks again for the support
>
> Jorge
>

Reply via email to