Memory grows up and not freed after big multi/exec

[GoogleCodeExporter]

What steps will reproduce the problem?

1. create one million hashes with 10 keys
2. save and restart the redis instance
3. load all keys with one multi/exec call

What is the expected output? What do you see instead?

The memory grows up very extensively, and is not freed when the exec is
done (but used_memory returns back to it's initial value, but not RSS)

What version of the product are you using? On what operating system?

Version 1.3.6 on Ubuntu server 9.10

Please provide any additional information below.

Attached two script in python using redis-py to create and read the data,
and a log file with, for each step, used_memory from INFO, and used_memory
from log file

Original issue reported on code.google.com by stephane.angel on 21 Mar 2010 at 12:22

Attachments:

• big_redis_create.py
• big_redis_read.py
• big_redis.log

[GoogleCodeExporter]

Here the conf used. Also notice that i used "export MMAP_THRESHOLD=4096".

Original comment by stephane.angel on 21 Mar 2010 at 12:36

Attachments:

• big_redis.conf

[GoogleCodeExporter]

The RSS still growing with small multi/execs but it's more discrete

Original comment by stephane.angel on 22 Mar 2010 at 10:52

[GoogleCodeExporter]

We figured this issue together, closing

Original comment by anti...@gmail.com on 27 Aug 2010 at 11:02

• Changed state: Invalid

[GoogleCodeExporter]

How was it solved?  I have the same problem. "info" says used_memory is 320meg.
But after a day of usage the RSS is at >700meg.

RSS is back to normal after restarting the server.

Original comment by nik...@gmail.com on 17 Sep 2010 at 4:54