As mentioned in my previous post, I wrote a patch for GlusterFS to increase its performance when operating on many smaller files. Someone told me the other day that this functionality has been pushed to the git repository. Would have been good to have heard about this sooner…

So all of you who emailed me positive feedback and asked to make it a tuneable in the translator config (thanks!) - please check out  the above link to the git repository.

On another note, it seems as if they’re breaking away from having the protocol version bound to the release version, good progress in my opinion!

I’ve had a closer look at glusterfs in the last few days following the release of version 2.0.1. We often get customers approaching us with web apps dealing with user generated content which needs to be uploaded. If you have two or more servers in a load balanced environment, you usually have a few options, an NFS/CIFS share on one of them (single point of failure - failover NFS is, well…), a SAN (expensive), MogileFS (good, but alas not application agnostic),  periodically rsync/tar | nc files between the nodes (messy, not application agnostic and slow), store files in a database (not ideal for a number of reasons). There are a few other approaches and combinations of the above, but neither is perfect. GlusterFS solves this. It’s fast, instant and redundant! 

I’ve got four machines set up, two acting as redundant servers. Since they’re effectively acting as a RAID 1, each write is done twice over the wire, but that’s kind of inevitable. They’re all connected in a private isolated gigabit network. When dealing with larger files (a la `cp yourfavouritedistro.iso /mnt/gluster`) the throughput is really good at around 20-25 MB/s leaving the client. CPU usage on the client doing the copy was in the realms of 20-25% on a dual core. Very good so far! 

Then I tried many frequent filesystem operations, untarring the 2.6.9 linux kernel from and onto the mount.  Not so brilliant! It took 23-24 minutes from start to finish. The 2.6.9 kernel contain 17477 files and the average size is just a few kilobytes. This is obviously a lot of smaller bursts of network traffic!

After seeing this, I dove into the source code to have a look, when I reached the socket code, I realised that the performance for smaller files would probably be improved by a lot if Nagle’s algorithm was disabled on the socket. Said and done, I added a few setsockopt()s and went to test. The kernel tree now extracted in 1m 20s!

Of course there’s always the drawback.. In this case it is that larger files take longer to transfer as the raw throughput is decreasing (kernel buffer is a lot faster than a cat5!). Copying a 620 MB ISO from local disk onto the mount takes 1.20 s with the vanilla version of GlusterFS, and 3m 34s with Nagle’s algorithm disabled. 

I’m not seeing any performance hit on sustained transfer of larger files, but at the moment I’m guessing I’m hitting another bottleneck before that becomes a problem, as it “in theory” should have a slight negative impact in this case.

If you want to have a look at it, you can find the patch here. Just download to the source directory and do patch -p1 < glusterfs-2.0.1-patch-erik.diff  and then proceed to build as normal.

Until I’ve done some more testing on it and received some feedback, I won’t bother making it a tuneable in the vol-file just in case it’d be wasted effort!

I attended the MySQL EMEA conference last thursday where I enjoyed a talk from Ivan Zoratti titled “Scaling Up, Scaling Out, Virtualization - What should you do with MySQL?”

They have changed their minds quite a bit. Virtualisation in production is no longer a solid no-no according to them (a lot of people would argue). Solaris containers, anyone?

As most of us know by now, MySQL struggles to utilise multiple cores efficiently. This has been the case for quite some time by now, and people like Google and Percona has grown tired of waiting for MySQL to fix it.

Sun decided to not go down the route of reviewing and accepting the patches, but are now suggesting - are you sitting down? - running multiple instances on the same hardware.
I’m not against this from a technical point of view as it currently actually does improve performance on multiple-core-multiple-disk systems (for an unpatched version) for some workloads, but the fact that they have gone to openly and officially suggest workarounds to their own problem rather than fixing the source of the problem is disturbing.

Granted, I suppose it makes sense to suggest larger boxes if you’ve been bought by a big-iron manufacturer. Also, I should be fair and note that Ivan at least didn’t say scaling out was a negative thing and that it’s still a good option.

If anyone asks me though, I think I’ll keep scaling outwards and use the more sensible version of MySQL

If you, like me, often work in a lot of terminals on a lot of servers, or even a lot of terminals on the same one, you may recognise the frustration of a lost bash history.
I don’t always gracefully log out of my sessions, so every so often my ~/.bash_history isn’t written and all my flashy commands are lost (the history buffer is only committed when you log out, everything that you see in `history` is not actually written to disk). I quite often find myself rewriting the same one-liners or long option list just because I closed my konsole or SecureCRT window without first logging out of all the sessions properly.

So I put some effort into finding a solution to this, and whilst reading through the bash manpage, I saw PROMPT_COMMAND. *pling*
export PROMPT_COMMAND='history -a'

To quote the manpage: “If set, the value is executed as a command prior to issuing each primary prompt.”
So every time my command has finished, it appends the unwritten history item to ~/.bash_history before displaying the prompt (only $PS1) again.

So after putting that line in /etc/bashrc I don’t have to find myself reinventing wheels or lose valuable seconds re-typing stuff just because I was lazy with my terminals.

This is one of those things that I should have done ages ago, but never took the time to.

Varnish has been able to provide caching for more than one backend for quite some time. The achilles heel with this has up until now been that it hasn’t been able to determine whether a backend is healthy or not. This is now a problem of the past! The backend health polling code is available in 2.0 beta1 Sadly it had a bug, so when using the ‘random’ director, it was unable to use the remaining healthy backend if all but one went MIA. I reported this bug and it was fixed in changeset r3174.

So as of now, you can safely use one varnish instance for several front-ends, thus eliminate double-caching (memory waste, unnecessary load on back-ends), reduce network traffic, do rudimentary load balancing, ease management etc.
With the obscene amount of traffic Varnish can push without putting a fairly basic system under any load worth mentioning, you can use a single front-end to serve several nodes in most setups.

Here’s an elementary sample VCL for how to do this:

backend node0 {
  .host = "127.0.0.1";
  .port = "80";
  .probe = {
           .url = "/";
           .timeout = 50 ms;
           .interval = 1s;
           .window = 10;
           .threshold = 8;

  }
}

backend node1 {
  .host = "10.0.0.2";
  .port = "80";
  .probe = {
#           .url = "/";
           .timeout = 100 ms;
           .interval = 1s;
           .window = 10;
           .threshold = 8;
        .request =
            "GET /healthcheck.php HTTP/1.1"
            "Host: 10.0.0.2"
            "Connection: close"
            "Accept-Encoding: foo/bar" ;
  }
}
director cl1 random {
    { .backend = node0; .weight = 1; }
    { .backend = node1; .weight = 1; }
}

#director cl1 round-robin {
#   { .backend = node1; }
#   { .backend = node0; }
#}

sub vcl_recv {
        set req.backend = cl1;
}

As you can see I’m defining the backends slightly differently. You need to define one of .url or .request, but not both for obvious reasons. If you go for the slighly simpler .url the default request looks like this:

GET / HTTP/1.1
Host: something
Connection: close

If this does not suit your need, comment out .url and use .request to roll your own. This aspect of Varnish is actually quite well documented, so I won’t repeat what’s on the trac page.

There is clearly a lot more you can and, more often than not, should do in the VCL than the above. This is a stripped down version which only pertains to the backend polling functionality.