Add information on how to enable DEBUG with the query engine

From: Keshav Murthy <keshav@couchbase.com>
Date: Tuesday, October 20, 2015 at 8:55 AM
To: Terry Dhariwal <Terry@couchbase.com>, Michael Nitschinger <Michael.Nitschinger@couchbase.com>, Query <query@couchbase.com>
Cc: Amarantha <amarantha.kulkarni@couchbase.com>
Subject: Re: N1QL questions

Hi Terry, below are the details.

Hi Amarantha, these should be in the docs…

Cheers, Keshav | 510.676.0824 | Blog | @rkeshavmurthy

$ curl http://localhost:8093/admin/settings -u Administrator:password

{"cpuprofile":"","debug":false,"keep-alive-length":16384,"loglevel":"INFO","max-parallelism":1,"memprofile":"","pipeline-cap":512,"request-size-cap":67108864,"scan-cap":0,"servicers":32,"timeout":0}

$ cat my.json

{"cpuprofile":"", "debug":false,"keep-alive-length":16384,"loglevel":"TRACE","max-parallelism":1,"memprofile":"","pipeline-cap":512,"request-size-cap":67108864,"scan-cap":0,"servicers":32,"timeout":24}

$ curl http://localhost:8093/admin/settings -u Administrator:password -XPOST -d@./my.json

{"cpuprofile":"","debug":false,"keep-alive-length":16384,"loglevel":"TRACE","max-parallelism":1,"memprofile":"","pipeline-cap":512,"request-size-cap":67108864,"scan-cap":0,"servicers":32,"timeout":24}

From: Terry Dhariwal <Terry@couchbase.com>
Date: Wednesday, October 14, 2015 at 9:15 AM
To: Root Murthy <keshav@couchbase.com>, Michael Nitschinger <Michael.Nitschinger@couchbase.com>, Query <query@couchbase.com>
Subject: Re: N1QL questions

Hi Keshav,

Is this client side or server side debug logging?

Can you point me to the documentation for enabling DEBUG with the query engine.
–

Always Scalable, Always On NoSQL

Terry Dhariwal
Solutions Engineer
+44.7460.877412
terry@couchbase.com

From: Keshav Murthy <keshav@couchbase.com>
Date: Wednesday, 14 October 2015 13:10
To: Tezza Tezza <terry@couchbase.com>, Michael Nitschinger <Michael.Nitschinger@couchbase.com>, Query <query@couchbase.com>
Subject: Re: N1QL questions

Hi Terry,

We don’t have an index-advisor tool yet.

The queries executed are not logged anywhere by default.
You need to enable the DEBUG log level for the query engine to log these.

Also,
Is there a customer situation or request for this? Please log a feature request with associated customer information. Thank you.

Cheers, Keshav | 510.676.0824 | Blog | @rkeshavmurthy

From: Terry Dhariwal <Terry@couchbase.com>
Date: Wednesday, October 14, 2015 at 4:56 AM
To: Michael Nitschinger <Michael.Nitschinger@couchbase.com>, Query <query@couchbase.com>, Root Murthy <keshav@couchbase.com>
Subject: Re: N1QL questions

Thanks team,

I'd like to expand on the question about identifying fields to index.

I'm aware of the explain command, but what I'm after is a way of identifying the most popular queries that are being executed. Is this logged somewhere?

What I'd like is a mechanism to identify popular queries and analyse them to ensure appropriate indexes are in place.

Sent from my HTC

----- Reply message -----
From: "Michael Nitschinger" <Michael.Nitschinger@couchbase.com>
To: "Terry Dhariwal" <Terry@couchbase.com>
Cc: "Query" <query@couchbase.com>, "Keshav Murthy" <keshav@couchbase.com>
Subject: N1QL questions
Date: Wed, Oct 14, 2015 12:39 PM

Hi Terry,

let me add to 1: the client SDKs now always set the query timeout by default, so we try to avoid that state explicitly.
Couchbase Logo
Always Scalable, Always On NoSQL

Michael Nitschinger
Senior Software Engineer, SDK Engineering
Skype: michael.nitschinger
Twitter: @daschl
Email: michael.nitschinger@couchbase.com

On 14 Oct 2015, at 13:35, Keshav Murthy <keshav@couchbase.com> wrote:

Hi Terry, answers in-line.

Cheers, Keshav | 510.676.0824 | Blog | @rkeshavmurthy

From: Terry Dhariwal <Terry@couchbase.com>
Date: Wednesday, October 14, 2015 at 1:47 AM
To: Query <query@couchbase.com>
Subject: N1QL questions

Hi Query Team,

I have a few questions that don’t seem to be included in our documentation:
How can we identify and kill long-running/run-away N1QL queries? We’re working on this feature for next release. Customers can set query timeout (which is not set by default). The index scan timeouts are automatically set to 2 minutes for secondary index scans.
How can we identify the most popular fields being used in where predicates that are NOT indexed? In simple set of queries, you can see the explain plan. Any query plans using primary scan means you don’t have appropriate index or query does not have any filters. For a deeper look, you’ll have to look at the index selection, spans created and the post scan filters in the explain.
When a data node is down, is there a way for the Query server to read/scan from replica data? Right now, Couchbase always reads from the master copy.
When you don’t provide an “order by” - how does the Query order the results? The result docs are not ordered in any predictable fashion. If customers expect ordering by some key, they should specify the respective ORDER BY clause. In future, we’d exploit the index ordering to avoid sorting for order-by when its logical.
Can we isolate specific Query servers to only run queries against certain buckets? Hmmm, I’m unaware of any techniques to do this…
Kind Regards,
Terry
–

Always Scalable, Always On NoSQL

Terry Dhariwal
Solutions Engineer
+44.7460.877412
terry@couchbase.com