<gchristensen>
znapzend could really use some introduced jitter
<gchristensen>
having several systems sync all an the some moment is not nice for the receiving end
<infinisil>
Yeah.. the longer I'm using znapzend the more things to improve I find
<gchristensen>
I'd really like it to operate on a systemd timer
<infinisil>
gchristensen: Btw: for set in $(zfs list -Hp -o name); do echo "zfs_snapshot_age_seconds{dataset=\"$set\"} $(zfs list -Hp -t snapshot -o creation -s creation "$set" | head -1)"; done
<infinisil>
But it doesn't quite do the same
<gchristensen>
neat
<gchristensen>
pretty close
<infinisil>
(And it takes longer because zfs list is called multiple times)
<gchristensen>
aye
<infinisil>
gchristensen: Or even fancier, `man zfs program`!
<gchristensen>
do it in lua or something? :P
<infinisil>
Yup!
<gchristensen>
so cool
<infinisil>
Yea, haven't used it though
<gchristensen>
wow, looking at the available lua api, I could really do some damage
waleee-cl has quit [Quit: Connection closed for inactivity]
<infinisil>
I just learned that Linux doesn't automatically put stuff from swap back into RAM when there's space
<infinisil>
I learned this because my swap is on an HDD, and apparently a lot of often-used things were in there..
<clever>
infinisil: but the reverse is true, there is a thing called swap cache
<clever>
$ grep SwapCached /proc/meminfo
<clever>
SwapCached: 932844 kB
<clever>
infinisil: when idle, linux will sometimes pre-write some stuff in ram to swap, but leave it in ram (and specially flag it read-only)
<clever>
if it needs the ram, it can insta-free the pages
<infinisil>
Neat
<clever>
if something writes to it, the pagefault (due to readonly) will alert linux, which invalidates the copy in swap, flags it r/w, and triggers a retry
<infinisil>
Hmm yes, some nice cache invalidation
<clever>
$ grep Cache /proc/meminfo
<clever>
Cached: 1600176 kB
<clever>
infinisil: i think this is both the read cache, and the mmap cache?
<clever>
the .text and .data sections of binaries, basically act like mini swap files
<clever>
and the kernel will read them in on-demand, but can insta-free them when needed
<clever>
$ grep Dirty /proc/meminfo
<clever>
Dirty: 556 kB
<clever>
infinisil: and this one, is the nasty one, any time you do things like ext2/3/4 on usb
<clever>
basically, the FS lets you have several gig of outstanding writes
<clever>
so it acts way faster then the disk can keep up
<clever>
and when you try to `sync` or `umount`, it will just hang for 10 or 20 minutes, while it writes it all to disk!!
<elvishjerricco>
infinisil, gchristensen: The odd thing about `zfs program` is that it put a lua interpreter in kernel mode, which seems like a bad idea. Dunno why they couldn't have just extended the ioctl API and written a user space library...
<infinisil>
clever: Earlier when I fixed my problem with swapoff I also had to wait like 10 minutes for it to copy everything away from swap :)
<elvishjerricco>
clever: gross :P
<clever>
infinisil: ive also found swapoff to have horid performance
<gchristensen>
elvishjerricco: that is surprising.
<clever>
elvishjerricco: does it at least need root to execute programs?
<elvishjerricco>
gchristensen: They justified it by saying there are other Turing complete drivers in the kernel that can sort of behave like an interpreter. Seemed like a weak argument to me
<elvishjerricco>
clever: No idea
<elvishjerricco>
Probably though
<clever>
elvishjerricco: i'm reminded of how windows renders fonts in-kernel, and fonts are now turing complete!
<elvishjerricco>
Dunno if delegated permissions have something for programs
<clever>
and there are font exploits, to gain ring0 on a remote box...
<eyJhb>
infinisil: the sad part was I didn't get any of the pizza.. :( I picked a friend up after she had hit the town, and I hoped she wouldn't be able to eat it all. I was very very wrong
<eyJhb>
Don't underestimate a hungry women I guess
__monty__ has joined #nixos-chat
<infinisil>
Haha
<eyJhb>
She took my cake as well.. :(
<infinisil>
Blimey!
drakonis has joined #nixos-chat
drakonis has quit [Quit: WeeChat 2.6]
drakonis has joined #nixos-chat
drakonis has quit [Quit: WeeChat 2.6]
ddima has quit [Ping timeout: 268 seconds]
endformationage has joined #nixos-chat
drakonis has joined #nixos-chat
ddima has joined #nixos-chat
<pie__>
so im gonna log into my github, havent been on in like two monts, lets see how many notifications i have
<pie__>
wow ok just two...
<pie__>
;_;
<lassulus>
github notifications are luckily quite sparse, except you opt in ffor them explicitly
<lassulus>
I miss prs to my own repos most of the time