Deploying CitusDB on AWS ends with an error

66 views
Skip to first unread message

maxim...@gmail.com

unread,
May 19, 2016, 2:04:52 PM5/19/16
to citus-users
Hello. 

I have a problem while trying to deploy CitusDB on AWS via teplates, that are published on citusdata.com.
I attach an image with an error.

Here are links for templates:

I tried different combinations of instances(t2, m3) and regions (California, Frankfurt, Tokio), result is always the same. 

Marco Slot

unread,
May 19, 2016, 3:37:33 PM5/19/16
to citus-users, maxim...@gmail.com
Thanks for reporting this, something seems to have changed about the way the CloudFormation init program works and I'm seeing a suspicious warning in the log. However, I am still able to start up a cluster using this CF template:

Is there any chance you could start a cluster with rollback disabled (in the "Advanced" section on the third screen of the CloudFormation start-up form) and post a copy of /tmp/logfile ?

maxim...@gmail.com

unread,
May 19, 2016, 4:52:23 PM5/19/16
to citus-users, maxim...@gmail.com
Thank you for a quick response, Marco.

There is no master created, I can connect to my worker nodes only, and here is what I've found in /tmp/logfile in both of them:

"FATAL:  could not map anonymous shared memory: Cannot allocate memory
 HINT:  This error usually means that PostgreSQL's request for a shared memory segment exceeded available memory, swap space, or huge pages. To reduce the request size (currently 1186930688 bytes), reduce  PostgreSQL's shared memory usage, perhaps by reducing shared_buffers or max_connections."

Through AWS I get the same error (from my first message's picture), status of the stack is CREATE_FAILED.

Which kind of instance did you use? 

четверг, 19 мая 2016 г., 22:37:33 UTC+3 пользователь Marco Slot написал:

Sai Srirampur

unread,
May 19, 2016, 5:49:51 PM5/19/16
to maxim...@gmail.com, citus-users
From the error it seems that postgres is trying to a allocate shared memory of 1186930688 bytes (around 1.19 GB) . This is because the default shared buffers in our cloud formation templates is 1GB. The memory allocation is failing, this may be because the instance is not having sufficient memory.

Could you please let us know what your exact instance types are and what memory they have ?

I could reproduce the same error on t2.micros (1GB Memory).

Sai

--
You received this message because you are subscribed to the Google Groups "citus-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to citus-users...@googlegroups.com.
To post to this group, send email to citus...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/citus-users/688820df-052f-449e-b8d0-7d41d000420f%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

maxim...@gmail.com

unread,
May 19, 2016, 5:59:01 PM5/19/16
to citus-users, maxim...@gmail.com
Sai, looks like it is exactly the reason. I remember that I went for m3-cluster and got an error message, which told that I tried to start too much instances up to that moment. 
So it could work for m3 correctly, actually. I'll check it a little bit later.

Thank you.   

пятница, 20 мая 2016 г., 0:49:51 UTC+3 пользователь Sai Srirampur написал:
Reply all
Reply to author
Forward
0 new messages