Analyzing the “elephant” together with colleagues

If your life as a PostgreSQL maintainer DBA is filled with questions “why so slow?” and “how to make the request not slow down?”, our service for analyzing and visualizing query plans explain.tensor.ru will make it a little easier due to engaging colleagues and updated tips

m / f "Koloboks lead the investigation"
m / f “Koloboks lead the investigation”

“Landon from the Capital of Great Britain”

Yes, at last what we have been asked for for the sake of our foreign colleagues has come true – we have supported the work of the service not only “in Russian”, but also in other languages. So far – only in English, but fully, including the translation of recommendation tips.

Discuss the problem plan wherever it suits you
Discuss the problem plan where it suits you

And so that a colleague can be quickly connected to the analysis of problems, the ability to quickly “share” link to plan via popular instant messengers, social networks or mail.

Plan hints

I already talked about the basic set of tips and how to use them in the article “Recipes for ailing SQL queries” – now we have made them even bigger and more convenient!

Recommended hints of plan nodes
Recommended hints of plan nodes
  • color and text coding of different types

    Now you can easily distinguish, at first glance, the problem of excessive filtering due to the lack of a suitable index from the waste of time on sorting that does not fit in memory.

  • several tips at one node

    An unsuccessfully used index with a large number of dropped records can easily coexist with irrelevant statistics on a table and be a consequence of it.

  • in the most conspicuous place

    You no longer need to scroll through a huge plan to see where the problem arose there – all the tips are already duplicated in the header, and just one click is enough to go to the desired node.

All tips - together, click - and you are there
All tips – together, click – and you are there

Likewise, clicking on a column header will immediately bring focus to node with maximum valueso find out “who filtered a million records here” can be done in half a second.

Scalable chart

If you see “hot Spots” on the flowchart, the execution of a large plan was difficult due to the resulting scrolling, but now it starts immediately zoom view… But you can return to the scrolling mode with one click.

Future battle map for performance
Future battle map for performance

Use it! If you have ideas or comments – please, in the comments.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *