User avatar
celeste_hearts_lesbianceleste_hearts_transceleste_hearts_demisexual flori_ava_star:~cursor_blinking @star@amazonawaws.com
1y
I didn't know that hosting a fedi instance required this many resources neocat_0_0
❤️2
2
0
3
2
User avatar
9/30 is the day :3 @theking@mskey.nekomimi.party
1y
@star huh, how many? I thought iceshrimp.net was light
1
0
0
0
User avatar
celeste_hearts_lesbianceleste_hearts_transceleste_hearts_demisexual flori_ava_star:~cursor_blinking @star@amazonawaws.com
1y
@theking I am running Iceshrimp.NET, and it already took my server down 3x today, just from CPU overload neocat_dizzy
❤️1
2
0
0
1
User avatar
9/30 is the day :3 @theking@mskey.nekomimi.party
1y
@star huh, weird, what CPU? How many cores allocated?
1
0
0
0
User avatar
celeste_hearts_lesbianceleste_hearts_transceleste_hearts_demisexual flori_ava_star:~cursor_blinking @star@amazonawaws.com
1y
@theking It was 8 reserved vCores on an AMD EPYC™ 7402P. I now limited it to 6
❤️1
1
0
0
1
User avatar
9/30 is the day :3 @theking@mskey.nekomimi.party
1y
@star that should be faster than my server???
1
0
0
0
User avatar
celeste_hearts_lesbianceleste_hearts_transceleste_hearts_demisexual flori_ava_star:~cursor_blinking @star@amazonawaws.com
1y
@theking This is the resource consumption spike that this very response you just made triggered:
columns from left to right are: CONTAINER ID, NAME, CPU percentage, MEM USAGE divided by LIMIT, MEM percentage, NET I/O, BLOCK I/O, and PIDS. The first container e58c1daec2c5 named iceshrimp.net_web shows high CPU usage at 222.68% in yellow text, using 342.8MiB out of 9.711GiB memory at 3.45%, network traffic of 45.6MB received and 32.5MB sent, block I/O of 55MB read and 0B written, with 40 process IDs. The second container 6e32ce9ef842 named nextcloud-app-1 shows 0.02% CPU, 297.9MiB memory usage at 3.00%, 1.16GB received and 97.2MB sent network traffic, 84.1GB read and 1.89GB written block I/O, with 14 PIDs. The third container 8db63c3fd47 named immich_server uses 0.23% CPU, 812.3MiB memory at 8.17%, 32.9MB network I/O in both directions, 204MB read and 1.67GB written block I/O, with 32 PIDs. The fourth container 8281ba2e409f named immich_machine_learning uses 0.45% CPU, 393.4MiB memory at 3.96%, 293MB received and 15.4MB sent network traffic, 79.7GB read and 781MB written block I/O, with 19 PIDs. The fifth container 90cf2440d212 named iceshrimp.net_db shows elevated CPU at 171.27% in yellow text, 146.1MiB memory at 1.47%, 469MB received and 730MB sent network traffic, 77.7GB read and 1.96GB written block I/O, with 24 PIDs.
❤️1
1
0
0
1
User avatar
9/30 is the day :3 @theking@mskey.nekomimi.party
1y
@star huh, that's really weird
I wonder if something is configured poorly be default for some hardware setups?

I thought iceshrimp net is supposed to be lightweight to an extent
1
0
0
0
User avatar
celeste_hearts_lesbianceleste_hearts_transceleste_hearts_demisexual flori_ava_star:~cursor_blinking @star@amazonawaws.com
1y
@theking Maybe I have to tone down the default configuration. Not sure, time will tell neocat_mug_owo
❤️1
1
0
0
1
User avatar
celeste_hearts_lesbianceleste_hearts_transceleste_hearts_demisexual flori_ava_star:~cursor_blinking @star@amazonawaws.com
1y
@theking neocat_notice neocat_notice neocat_notice neocat_notice Sometimes, PEBCAK
screenshot of a configuration file with a dark gray background and white monospaced text, showing backfill settings for a social media platform. The file contains two main sections labeled Backfill:Replies and Backfill:User in white text within square brackets. The Backfill:Replies section includes several lines of comments starting with semicolons explaining that this feature enables backfilling of replies but is disabled by default due to performance concerns. The comments note it is experimental and requires remote instances to expose a replies collection. The setting Enabled equals true is shown. Additional settings include FetchAsUser equals true, which identifies the user requesting replies instead of using an anonymous actor, NewNoteDelay equals 5m for delaying reply backfilling on new notes, and RefreshAfter equals 15m for the cooldown between multiple backfill attempts. The Backfill:User section has similar warnings about enabling backfilling of user posts on follow, noting performance impacts and experimental status. It requires remote instances to expose an outbox collection. This section also shows Enabled equals true.
❤️1
1
0
1
1
User avatar
celeste_hearts_lesbianceleste_hearts_transceleste_hearts_demisexual flori_ava_star:~cursor_blinking @star@amazonawaws.com
1y
Let's see if disabling it helps! I thought that this would only apply to posts from other people, though. Hm :3
❤️1
0
0
1
1