Implementing better caching technique

After playing a little bit with memcache I had to implement better caching technique. Site was down for some hours, I look what happened…it was an error, which I didn’t see any time before…

Problem was with session table – it is stored in mysql. In log I saw something is not goot with session table, so I tried to repair it. After myisamchk I saw something is wrong with inodes…First I checked, if hdd is not full – no, there is more than 100 GB of free space. I am not unix guru, so I didn’t know what exactly is inode. Google helps again, so I checked free inodes with df -i, and yes – no inodes was free. For you, who doesn’t know what inode means – read more at wikipedia. Problem is – current cache implementation cached into files – in dir structure like this: cache/af/1b/4c/2d/af1b4c2d…gz – on hdd was so many files, no inode was free to make another file. This was serious problem, so I deleted some portion of cache and start thinking of better implementation, because this one has own fallbacks – for example deleting whole cache takes 1 week or so…strange,eh ?
So I decided to implement cache into MySQL, it has own drawbacks too, but it should work better in general. We will see, if it was good decision 🙂