manton
manton

The weakest link in Micro.blog’s server infrastructure remains our two Redis servers, which often get up to the breaking point of memory usage. That blew up today and I just finished restoring it. 48 GB memory is not enough, but I’m hesitant to do another upgrade… Need to trim the bloat instead.

|
Embed
Progress spinner
In reply to
jim
jim

@manton Bummer and thanks for quick action. Presuming some small level of data loss?

|
Embed
Progress spinner
umerez
umerez

@manton I wish I could be of any help. No clue. A lawyer in Spain of very little use. I just want to thank you for your work.

|
Embed
Progress spinner
manton
manton

@estebantxo Thanks for the support!

|
Embed
Progress spinner
manton
manton

@jimmitchell A little, but mostly a cache that can be restored. If anything is missing, it should fix itself.

|
Embed
Progress spinner
kottkrig
kottkrig

@manton That sounds like a real challenge, especially when memory limits are being pushed so hard. I wonder how the new Solid Cache (introduced in Rails 8) holds up under such load. Using the disk instead of memory sounds like a fair trade off in this case.

|
Embed
Progress spinner
lukemperez
lukemperez

@manton Remarkable though that it is only 48 that keeps this place running. I would have a higher number if asked to guess.

|
Embed
Progress spinner
thisismissem@hachyderm.io
thisismissem@hachyderm.io

@manton what's being stored in redis?

|
Embed
Progress spinner
manton
manton

@thisismissem Probably too many things. Lots of little caches, but primarily it’s used to power everything in the timeline with sorted sets.

|
Embed
Progress spinner