01:03:28 ty ek have a fun day 01:08:02 tuaris: Try rquery 01:08:13 I dunno if it has the query option 12:25:38 in the freebsd installer, the auto zfs screen, pool options are -O compress=lz4 -O atime=off, but isn't the property compression not compress? 12:57:29 kerneldove: try issuing `zfs get compress ` or `zfs set compress=off ` 12:58:44 it is probably better to spell it out entirely, but it works either way 13:00:06 oh right 13:00:26 kerneldove: in zfsprops(7): This property can also be referred to by its shortened column name compress. 13:01:22 we should probably use the canonical name in the installer though 13:07:33 ... was about to submit a review for that but i think phab just went down for maintenance 13:28:09 omg zfsprops. i looked everywhere for the info. man zfs, man zpool, man zpool-create, man zpoolprops 13:28:33 then did a web search for it and only saw compression not compress 13:29:24 https://reviews.freebsd.org/D51572 13:29:28 compression woulda been better to me as a new user, and an experienced user would be fine with either, so i'd use compression imho 13:30:04 tyvm 13:32:13 so i used compress=off during install, went through install then update and reboot, now sudo top shows 43M compressed in the ARC section. but i turned it off? 13:32:27 i confirmed it's off with zfs get compression 13:35:32 ARC compression is something diffrent then the compression of the datasets 13:36:17 oh 13:37:03 you only control compression of the datasets (on disk) with zfs set compression 13:37:29 ARC is in RAM only 13:38:09 no way to disable ARC? 13:38:49 zfs set primarycache=none, but you almost certainly don't want to do that except in very specific workloads 13:39:35 ok i'll try with it on i guess. i wanna jam pack RAM with workload so trying to limit any other uses of it 13:40:13 memory used by ARC is automatically released when something else needs it, also in that case you want more compression, since that means it uses less memory to store the same data 13:40:19 i'm gonna fill ram to 105% (5% of swap used) and crank up that up until vmstat pi po starts showing regular numbers in the double digits 13:41:33 generally you turn ARC off if you got less than 4 GB of RAM 13:42:38 Remilia: i would really recommend just limiting the size there (by setting arc_max) rather than turning it off, the system isn't designed to run without disk caching... 13:42:38 I guess you could also do it if you run ZFS on a memory-backed virtual device? 13:43:31 ivy: I agree, just that iirc the installer disables it automatically on that <4 GB condition 13:43:42 or maybe I am misremembering 13:53:32 my concern with ARC is i don't see it helping with my workload. lots of different files being accessed, and rarely is the same file accessed repeatedly, so i can see it just using ram for no upside 13:53:36 am i wrong? 14:00:33 kerneldove: imagine an application that reads from a file in 4kb chunks. now imagine that file in stored in a 128kb zfs record. with arc, the record will be read once, then the application's 4kb reads will be served from arc. without arc, reading the 128kb record would require reading the same block disk from 32 times 14:00:58 s/block disk from/block from disk/ 14:01:29 but if every file is only read once is that any benefit? 14:01:47 yes, because reading the 128kb record *once* still requires reading it from disk 32 times in this scenario 14:01:59 my workload is many files with each read and served very few times 14:02:09 ok 14:02:16 application reads 4kb -> zfs reads the 128kb record -> returns 4kb of it to the application -> discards the block (because arc is disabled), the application reads the next 4kb block, zfs has to read the entire record from disk again... 14:02:47 generally, primarycache=none is appropriate where a) the application has its own caching system (e.g., mysql, oracle), and b) the zfs recordsize matches the application block size 14:02:50 ah i think i understand better 14:03:01 ok ty 14:03:04 in that case, the application will always read (and cache) an entire record and there's no benefit to caching it again in arc 14:05:02 but you can always test it with your own workload and see what works better 16:44:26 ah, found in FreeBSD-main-amd64-gcc14_build - Build #1097 16:44:32 wrong channel 20:16:21 good suggestions on a VPS provider for FreeBSD? my old one still runs on Linode, but I'm worried that akamai or whoever will break it at some point 20:17:10 mmlj4: I use vultr 20:17:26 https://www.vultr.com/ 20:18:57 I'm quite satisfied with the service, but I never tried anything else so far, so my judgement can't be objective. 20:27:38 thanks 20:28:16 you're welcome :) 20:43:31 I got a vultr account a year back, they gave me grief about email, so I bailed 20:43:33 grr... 20:56:13 grief about email? 22:10:03 Probably blocked port 25 or something. 23:04:20 Can anyone who uses ip6addrctl(8) (or is better than me at reading source code) say whether its use of precedence conforms to RFC 6724 section 2.1 (that is, a prefix with higher precedence is more preferred)? 23:20:12 dacav: what TommyC said... and they refused to remove the block, so I bailed because the entire reason for spinning up the VM was as a mail host 23:20:45 pity, I'd heard good things about vultr too