Oddbean new post about | logout
▲ ▼
sommerfeld | 5 months ago (raw) | export | reply | flag +3 
 Fulcrum DB got corrupted, AGAIN. fml 💀🤬🪦
▲ ▼
greenart7c3 | 5 months ago (raw) | root | parent | reply | flag 
 It's a pain to sync a fulcrum db
▲ ▼
sommerfeld | 5 months ago (raw) | root | parent | reply | flag 
 I have a secret: I always sync on my desktop and then copy it to my node. Not only is it much faster, I also always have a backup on my desktop up to the last time I did this, so I never have to do it from scratch.
▲ ▼
greenart7c3 | 5 months ago (raw) | root | parent | reply | flag 
 Nice. I'll start doing it next time I need
▲ ▼
BlackCoffee | 5 months ago (raw) | root | parent | reply | flag 
 Do you know the cause? I got a UPS for my dojo but still haven't gone back to Fulcrum.
▲ ▼
greenart7c3 | 5 months ago (raw) | root | parent | reply | flag 
 For me it was power loss and one time the SSD was overheating
▲ ▼
sommerfeld | 5 months ago (raw) | root | parent | reply | flag 
 I honestly don't know. This happened by just rebooting...
My theory is that when you reboot, it kills all processes "nicely" with SIGTERM but if they don't stop after some time, it kills them forcefully with SIGKILL. I think Fulcrum can take a while to shutdown if it gets killed while writing to the DB, so when it gets SIGKILLed later, DB gets corrupted.
▲ ▼
cloud fodder | 5 months ago (raw) | root | parent | reply | flag 
 Do you have timeoutstopseconds set? I guess I should ask if you're using a systemd unit first 😁

It should wait for it on a reboot if you have the settings set and it's running in the system.slice
▲ ▼
c2e5c5cb | 5 months ago (raw) | root | parent | reply | flag 
 Use the latest version. The problem is not Fulcrum, but the underlying rocksdb, which was very shaky <v8. Now it’s really stable, I love Fulcrum. It’s great software. 

However, I regularly make rsync backups of the database directory, just in case. But I do that with the Bitcoin blockchain as well because I don’t want to wait weeks for it to resync/reindex.