Any desire for updated MogileFS docs?

Arthur Bebak abebak at
Thu Jul 12 18:43:54 UTC 2007

Earl Cahill wrote:
>> 1.  Would people find real value in updated docs?

Beyond just a straight install guide I think there's a crying
need for an operational admin user's guide to Mogile. If you're running
the system in production (and many people do) here's just a
smattering of the kind of questions you need to know:

What are the best practices for tuning the system?
How do you deal with failing drives/trackers/storage nodes?
How do you do recovery?
How do you setup different replication policies (with examples)?
What does the fsck option do behind the scenes? When do you use it?
What problem does the slave node solve and how do you work with it?
What's the normal list of processes which runs on trackers/storage nodes?
Are there sample startup/shutdown scripts for various architectures?
What kind of monitoring should be done to measure the health of the system?
How exactly do you get detailed statistics out of the system (beyond mogadm stats)?
How do you debug the system?
What are the metadata storage features and how and when do you use them?
What's an example setup (i.e. show the config files!) with Perlbal as a proxy?
How do you measure performance of the system?
How do you thoroughly test the functionality of the system?
How do you stress test the system?

I'm sure there are others. Some of these I've seen asked on the
mailing list, many I have not but I assume people running production
servers have the answers.

This is either fodder for a FAQ or something a bit more organized and kept
up to date.

And before you ask, I certainly don't have all the answers :)
Arthur Bebak
abebak at

More information about the mogilefs mailing list