Basic monitoring should be done on a regular basis. This keeps you informed as to the general condition of your server and may warn of impending problems.
This article concentrates on memory management using the command 'free'.
Memory management is actually very easy to do. What you do about any problem is a different matter and we'll look at monitoring what is actually using the memory and how much it's using in a later article.
To start with, log into your server and give the 'free' command:
The output consists of four lines:
Well, for a start, that's all a bit confusing as the answer is in Kb. I don't know about you but dividing the reported figures by 1024 is a bit of a pain.
Let's make it more readable by adding the '-m' option which will displays the output in megabytes:
One of the aliases in my .bash_profile is:
alias free="free -m"
So I only have to enter 'free' to get a readable output.
Buffers and Cache
The first line gives the column headers: total, used, free and so on.
A common mistake is to read the second line and have a panic attack as that line reports:
. total used free shared buffers cached Mem: 254 249 5 0 15 181
However, the figure of 249MB out of 254MB includes cached and buffered memory.
Unless you are particularly interested in the buffers and cached figures, ignore this line.
Concentrate on the third line:
. total used free shared buffers cached Mem: 254 249 5 0 15 181 -/+ buffers/cache: 51 202
This tells us what memory is being used and is not available for another process.
So it reports 51MB of 254MB being used by various tasks and applications. I have 202MB to play with and for my application to grow into.
Not bad, but then I don't have anything installed on this test slice - this was just after a base install and update.
The final line tells me how much swap space I am using:
. total used free shared buffers cached Mem: 254 249 5 0 15 181 -/+ buffers/cache: 51 202 Swap: 511 0 511
As you can see, I have 511MB of swap available and have used none of it.
If you are using swap space then you need to investigate what is using so much memory that it had to resort to hard drive swapping to work. Often it's a case of your application 'outgrowing' your Slice and you may need to upgrade it to include more RAM.
The odd bit of swapping is fine as it is, after all, there to be used. However, consistent swapping or very high levels of swapping does mean your Slice is struggling and it will make your web application slow down.
Actually, that's a good example as to why regular monitoring of your Slice is a good idea - your site may be up and running but it may be struggling and becoming slower and slower.
It's a lot better to catch any issues before they result in downtime and a non-operational site.
That's it - quick and easy. Next time we'll look at system monitoring with 'top' which will pinpoint what processes are actually using the memory and, more importantly, how much each process is using.