You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@drill.apache.org by Алексей Тришин <al...@gmail.com> on 2018/07/18 05:29:49 UTC

Apache Drill and PostgreSQL

Hello, I am Alex and I recentrly started to use apache drill. One of
connected to apache drill databases is postgresql. I dumped 65gb of csv
files there and tryed to select something from there via apache drill like
"select imei from billing limit 10". It starts perform this query for a
while and then i get an error:
"Error: CONNECTION ERROR: Connection /172.17.0.16:53996 <--> 986910249f08/
172.17.0.16:31010 (user client) closed unexpectedly. Drillbit down?
[Error Id: 552dedfa-7ebb-42d7-bedd-57a599f78bbc ] (state=,code=0)"

And when I add some where clause like ... from billing WHERE imei =
'123456789', drill responses me after 80 seconds and gives me expected
result. The same queries in PrestoDB work fine and much more fast. What do
I do wrong?

Thank you, Trishin Alex.

Re: Apache Drill and PostgreSQL

Posted by Kunal Khatua <ku...@apache.org>.
Hi Alex

Do you have a lot of files in your 65GB of CSV dump, and are the rows very wide? Is the error instantaneous or does it take a while.
The error for your example would typically occur if the Drillbit is very busy doing something, resulting in a timeout or lack of a heartbeat from the Drillbit.

For the 2 queries you ran, what do the query profiles say about where the most query time was spent and what was the planning time?

~ Kunal
On 7/18/2018 7:52:20 AM, Алексей Тришин <al...@gmail.com> wrote:
Hello, I am Alex and I recentrly started to use apache drill. One of
connected to apache drill databases is postgresql. I dumped 65gb of csv
files there and tryed to select something from there via apache drill like
"select imei from billing limit 10". It starts perform this query for a
while and then i get an error:
"Error: CONNECTION ERROR: Connection /172.17.0.16:53996 <--> 986910249f08/
172.17.0.16:31010 (user client) closed unexpectedly. Drillbit down?
[Error Id: 552dedfa-7ebb-42d7-bedd-57a599f78bbc ] (state=,code=0)"

And when I add some where clause like ... from billing WHERE imei =
'123456789', drill responses me after 80 seconds and gives me expected
result. The same queries in PrestoDB work fine and much more fast. What do
I do wrong?

Thank you, Trishin Alex.