[Postgres-xl-general] Invalid memory alloc request size

Aaron Jackson ajackson at revionics.com
Thu Sep 18 22:19:49 PDT 2014


I've looked at this one for a while, so I will throw it out here and continue looking to see if I can make sense of it.

I have an i2 with 32GB of memory.  Kernel parameters are set as follows:
kernel.shmall = 7864320
kernel.shmmax = 32212254720
kernel.shmmni = 4096
This machine is host to one coordinator and two datanodes (datanode1 and datanode4).  Datanode4 was recently added and I wanted to redistribute data that was currently on 1,2 and 3 onto 4.  Thus, I issued the following command.
ALTER TABLE client.foo ADD NODE (datanode4);
It processes the request, eventually failing with the following:
ERROR:  invalid memory alloc request size 1073741824
There doesn't appear to be a clear point of origin of the failure; both datanode1 and datanode2 appear to have been executing a COPY at the time of the failure.  Though the only error message in a log appears to have occurred in the coordinator itself.

Here are some memory settings if they help, remember that datanode1 and datanode4 are on the same instance.
coord1
work_mem = 144MB
maintenance_work_mem = 1GB
effective_cache_size = 2GB
shared_buffers = 1GB

datanode1
effective_cache_size = 6GB
work_mem = 144MB
wal_buffers = 16MB
checkpoint_segments = 32
shared_buffers = 3GB

datanode4
effective_cache_size = 6GB
work_mem = 144MB
wal_buffers = 16MB
checkpoint_segments = 32
shared_buffers = 3GB
For completeness sake, I will check the other two datanodes to make sure it's not one of them that was failing.

Aaron
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.postgres-xl.org/private.cgi/postgres-xl-general-postgres-xl.org/attachments/20140919/663a2b28/attachment.htm>


More information about the postgres-xl-general mailing list