Server runs out of RAM – How to find cause

memory

My website seems to go down for up to an hour almost every day. My host says it's running out of RAM. I don't know much about servers so would appreciate advice on what to do next.

The website (3dsbuzz.com) has a WordPress section, Wiki, Gallery and a VBulletin forum each with their own custom code and plug-ins. So there are lots of potential places which could be inefficiently coded. It runs from a cloud server with 2GB of RAM. The MySQL database is 370MB. We get around 30,000 page views per day.

What would be the best way to reduce the amount of down time? Should I upgrade my server (I can't really afford to) or is 2GB reasonable? I have plenty of errors in my error log, but they don't always occur around the same times as the server going down, so I'm not sure how relevant it is.

Best Answer

2GB of RAM looks ok.

Some clues to check:

Database
- Configure mysql to get slowlog - to find bottlenecks.
- Check mysql engine.
  - If there is MyISAM - it could be problem because of whole table locks with writes (for example VBulletin could cause many writes).
  - If this is the problem - convert every table not using FULLTEXT index (or at least often changed tables) to InnoDb (or even better XtraDB from MariaDB or Percona)
Web serwer
- Biggest problem could be apache with mod_php. (especially if because of mysql it top time requests beginning to lasts longer).
  - If you have this:
    - preferable switch to nginx.
    - If you cant - use fcgi php (preferable php5-fpm).

And consider hiring professional who can audit situation and help with configuration.

Related Solutions

SQL Server 2005 – Find Database RAM Usage

It was most likely caused by a query wanting to read more pages into the buffer pool, and the buffer pool grabbing more memory to accomodate that. This is how SQL Server is supposed to work. If the box experiences memory pressure, it will ask SQL Server to give up some memory, which it will do. The customer shouldn't be concerned.

You can use the DMV sys.dm_os_buffer_descriptors to see how much of the buffer pool memory is being used by which database. This snippet will tell you how many clean and dirty (modified since last checkpoint or read from disk) pages from each database are in the buffer pool. You can modify further.

SELECT
   (CASE WHEN ([is_modified] = 1) THEN 'Dirty' ELSE 'Clean' END) AS 'Page State',
   (CASE WHEN ([database_id] = 32767) THEN 'Resource Database' ELSE DB_NAME (database_id) END) AS 'Database Name',
   COUNT (*) AS 'Page Count'
FROM sys.dm_os_buffer_descriptors
   GROUP BY [database_id], [is_modified]
   ORDER BY [database_id], [is_modified];
GO

I explain this a little more in this blog post Inside the Storage Engine: What's in the buffer pool?

You could also checkout KB 907877 (How to use the DBCC MEMORYSTATUS command to monitor memory usage on SQL Server 2005) which will give you an idea of the breakdown of the rest of SQL Server's memory usage (but not per-database).

Hope this helps!

Linux – “nice” for memory management

You can use cgroups to do this. It lets you set limits for memory and memory+swap, so you can force processes into swap.

Alternatively it also lets you adjust the swappiness of each cgroup independently, so you can increase the odds that your background pages will be swapped out:

https://en.wikipedia.org/wiki/Cgroups

Best Answer

Related Solutions

SQL Server 2005 – Find Database RAM Usage

Linux – “nice” for memory management

Related Topic