After Samba proved too problematic for Geograph we switched to NFS and have enjoyed a good period of stability ever since. The webservers are barely breaking a sweat, but the load on our database is a little higher that I hoped. By no means maxed out, but definitely a lot more than it needs to be. The next week or so will be spent distributing some of that load to the webservers (like using memcached for session storage) and also looking at some of the heavyweight queries and tasks to see how they could be performed more efficiently….
Everybody is linking to Bruce Schneier Facts today. As a long-time reader of his excellent Crypto-gram newsletter, it tickled me. Unless you are into cryptography, your mileage may vary. I even added a few of my own!
My favorite is “Bruce Schneier writes his books and essays by generating random alphanumeric text of an appropriate length and then decrypting it.”
Can’t write too much as I’ve been on holiday for a week and I’ve lots to catch up on, but last night the new Geograph servers went live! At the appointed time we sync’d the database and images onto the new servers, then simply tweaked the haproxy configuration, opening the floodgates!
We’ve a couple of minor teething problems but everything seems to be holding up well!
The only major issue is every now and then the smbiod process on one of the webservers consumes 100% of the CPU, taking it out of the cluster. Not seen this issue before, will investigate it during the week. I thought Samba would be a low-maintenance solution for file sharing between servers, as I’ve never had a good time with NFS….we shall see!