I’m coming close to finally getting my iOS app Tubulate deployed to the app store. Part of that process is to create a web site to promote it and to serve as a contact point for the app store page.
I decided to use a free Openshift gear to host my site and Node.js / Express to implement it. Openshift comes with a default disk quota of 1 Gigabyte, enough I figured, for anything I could ever need for such a small site.
However I decided to get funky and learn more about node, express, sessions etc and to use a MongoDB database to bootstrap a user login system (Can you see why this app has taken ages to get published yet?, I do call my blog overtaken by events for a reason…)
So I code up the implementation of sign-in, register and sign out locally and all is golden. I push the local repository out to Openshift and all works as expected. I make a minor modification after the initial push and I redeploy. Eek! There’s a warning message at the end of the deploy saying that I was at 99% of my disk quota.
Time to investigate. A quick look at the quota’s from the rhc command shows the problem (tubular is my OpenShift App name)
> rhc app show tubular --gears quota
Gear Cartridges Used Limit
------------------------ ----------------------- ------ -----
54b11fa5e0b8cd5ac0000172 nodejs-0.10 mongodb-2.4 0.9 GB 1 GB
That’s a little on the excessive size considering that the database has exactly two users in it and nothing else. I try a basic tidy up
> rhc app tidy tubular
But that didn’t really have any noticeable effect. Logging on to the server and enjoying a little du action led me to the mongodb folder as being the main culprit:
> rhc ssh $@ -a tubular
Connecting to ...
Find where MongoDB stores its data
> cd ~/mongodb/data/
Check what’s going on
> du -mH .
1 ./tubular
385 ./journal
481 .
Total space used is 481 MB with 385 of those being the journal. A quick look at the folder content shows the actual tables are 34MB each (data table and ns file), hardly ideal for a tiny app but acceptable.
> ls -l --block-size=MB
total 101MB
-rw-------. 17MB Feb 1 07:11 admin.0
-rw-------. 17MB Feb 1 07:11 admin.ns
drwxr-xr-x. 1MB Feb 1 06:20 journal
-rw-------. 17MB Feb 1 06:35 local.0
-rw-------. 17MB Feb 1 06:35 local.ns
-rwxr-xr-x. 1MB Feb 1 06:35 mongod.lock
drwxr-xr-x. 1MB Feb 1 06:15 tubular
-rw-------. 17MB Feb 1 06:20 tubular.0
-rw-------. 17MB Feb 1 06:20 tubular.ns
The journal however is taking up 385MB (38.5% of the quota for the gear!) So something needs to be done.
The journal is used to cache updates to the table before they are written and provide a recovery mechanism from MongoDB crashes or unexpected power outages mid-write. Journals are a good thing™, but not compulsory , by default they’re not even enabled for 32 bit servers. So ours has to go! (But think long and hard before you do this for a write heavy site).
To disable journaling, edit the MongoDB configuraion file mongodb.conf.
> vim ~/mongodb/conf/mongodb.conf
Add the line
nojournal = true
Then restart your MongoDB cartridge (the line below assumes you are still logged in via ssh, to do this locally on your machine use rhc)
> ctl_app restart
Cart to restart?
1. nodejs-0.10
2. mongodb-2.4
Enter 2 (your listing will vary, your number may be different. I’m patronizing you now aren’t I… )
Stopping MongoDB cartridge
Starting MongoDB cartridge
Waiting for mongo to start...
Finally the payoff, we remove the logs
> cd ~/mongodb/data/journal/
> ls -l
total 393228
-rw-------. 134217728 Feb 1 06:20 prealloc.0
-rw-------. 134217728 Dec 8 13:19 prealloc.1
-rw-------. 134217728 Dec 8 13:19 prealloc.2
> rm *
Back on my machine
> rhc app show tubular --gears quota
Gear Cartridges Used Limit
------------------------ ----------------------- ------ -----
54b11fa5e0b8cd5ac0000172 nodejs-0.10 mongodb-2.4 0.5 GB 1 GB
Isn’t that better :-)