So the costs for running my very-light-use ceramic node spiked from $0 in august to nearly $2000 in September. The vast majority of this is write to storage costs - even a low cost-per-write skyrockets when there are millions of them.
I haven’t done any work with ceramic over this period, so I’m not sure what has happened. There would have been (max) of about 20-50 writes to my apps streams (to the best of my knowledge).
Any idea what could be happening? I’ve had to turn off our Ceramic integration until I can figure out what the hell is going on.
Are your ceramic daemon logs being written to storage?
I found my boot drive where my Ceramic logs were written to was rapidly filling up with 10GB worth of logs every week under .ceramic/logs/http-access.log and .ceramic/logs/pubsub.log and causing a 502 error (Note: Please don’t brick the node if I can’t write logs)
My ceramic daemon has written 12 logs (6 each) every second since my logs start on September 5th.
So this seems to be the culprit (although I’ll have to wait a few days to get confirmation). The pubsub was writing 6+ lines per second. It seemed to be storing every request that came in.
In my case, I simply disabled log-to-files. Azure stores all your container output anyway, so all the interesting information (startup/config/error messages) are logged automatically.
I also tried raising the --log-level from 2 (important) to 3 (error) via cmd-line argument, but the container refused to start with