|
echinopsis posted:imma keep being nice and generous until its no longer needed or wanted
|
![]() |
|
![]()
|
# ? Jun 9, 2024 08:16 |
|
Sagebrush posted:this is a legal and mental distinction, not an emotional one we have kids together dude, that we want to be well involved with anthonypants posted:why do you believe that are you able to assess whether it is needed or wanted communication for example there is a poo poo load of yard work around this property whereas I have none. we agree that we want the kinds to have a backyard and stuff. so I say "feel free to ask for me help to do some gardening or taking green waste to the dump thjis weekend" our kids will se that we are still able to be produtive adukts about all of this like maybe yhoure all right, but I just dont really see that the problem is. help until its no longer needed or wantd. im too lazy to martyr muself
|
![]() |
![]()
|
|
![]() |
|
i moved all my stuff to gandi. its good
|
![]() |
|
all the domains i never use except as mail forwarders are lovingly cradled in their arms
|
![]() |
|
echinopsis posted:we have kids together dude, that we want to be well involved with Maybe you should dig a well.
|
![]() |
|
spankmeister posted:Maybe you should bring her a printer
|
![]() |
|
gently caress. I have a home linux server that maybe a year or so ago, I decided I would make a backup of its boot disk by just dd cloning it. This worked fine, and I could readily boot from the alternate disk and everything would function as normal. Also about a year ago maybe a bit less than that, I first encounter a problem with the server that it fails to boot some percent of the time - I never actually calculated what percent because I would stop rebooting it after it had a successful boot. It would always come on after a few tries though, and it being on 24/7, this didn't really matter too much. So today there was a power outage and I finally got frustrated enough to try to debug the problem. I actually had to hook up a null modem/USB<->serial cable to debug it because it would basically completely poo poo the bed when it failed. In any case after like 2 hours, it turned out that even though bios was booting the correct disk, linux was mounting the root partition from the backup partition, which had old kernel modules that wouldn't load with the newer kernel that was present on the main disk (because this is arch linux and there's a new kernel like every week). This was happening because /dev/sda (the backup) and /dev/sdd (the main) were being assigned inconsistently at startup - the main was coming up as /dev/sda and vice-versa. So I guess what I'm saying is, if your OS is indeed a piece of poo poo, make sure you're at least using UUIDs in your /etc/fstab. Edit: Actually lmao I don't know what the gently caress. I thought it was the issue because it booted the first time after changing it. But I must have changed the wrong disk because the active fstab is using /dev/sda And the best part is /dev/sdd is not present in /dev Edit2: the backup was actually showing up at /dev/sde Also, the UUIDs on the partitions were the same on both disks because I used dd to do the backup so even just fixing fstab wouldn't have fixed it. Colonel Taint fucked around with this message at 03:07 on Jan 13, 2018 |
![]() |
|
no comp[uters, no problesm
|
![]() |
|
Colonel Taint posted:gently caress. a dd backup is trash unless the disk was un-mounted while you read it
|
![]() |
|
if you want a crash consistent backup on Linux step one is to create an lvm snapshot doesn’t matter what backup tool you are using. step one is the snapshot.
|
![]() |
|
i think the main problem is that cloning a disk also clones the uuid, so who the gently caress knowns what disk linux will mount if they look virtually identical
|
![]() |
|
sounds like op has no idea what their doing
|
![]() |
|
spankmeister posted:sounds like op has no idea what their doing hell, same
|
![]() |
|
Tankakern posted:i think the main problem is that cloning a disk also clones the uuid, so who the gently caress knowns what disk linux will mount if they look virtually identical yes if you are using software raid-1 or doing dd backups, you can't mount by uuid or label anymore, for obvious reasons
|
![]() |
|
colonel taint's larger problem is that his backup strategy cannot work. that backup volume is probably corrupt as hell. he is fortunate the boots failed!
|
![]() |
|
I mean what is the actual problem with that though? Aside from confusion from the duplicated uuids. The backup actually booted and functioned fine when I tested it originally - it technically functioned as designed even when the boots were failing, just the kernel modules weren't matching the kernel version and a basically nothing could load because of that.The server doesn't really do much of anything and pretty much everything written (except for system updates and logs) is on a different mdm array. I'd sync the system before doing dd so I would at most expect maybe some messed up logs. I realize it's generally "bad practice" and if this was something super important or someone else's server, I would probably just blindly take the advice but as it is it's just easy to sync and dd the thing now and then. This is admittedly a dumb hill to die on, but I don't know. Enlighten me.
|
![]() |
|
literally any filesystem data OR metadata change will be a dangerous write that could corrupt the filesystem. changes can remain un-flushed for days or weeks on a linux system. you made a copy of whatever happened to live on disk, and each block comes from a different time domain you could epically hose your poo poo, and not notice until it's too late. (most notably, the backup disk should NOT have booted cleanly -- it should have dumped to single user because it's un-clean with errors that require manual intervention. booting successfully, in this case, is a sign of pervasive corruption. oops.) either do your backup offline (i.e. from a livecd) or take an lvm snapshot
|
![]() |
|
like obviously nobody cares if a few blocks of log mesages in the middle of a file are corrupt in the backup. that's no big loss having the directory trees corrupted, on the other hand, is really bad
|
![]() |
|
Notorious b.s.d. posted:like obviously nobody cares if a few blocks of log mesages in the middle of a file are corrupt in the backup. that's no big loss looking at my own post we have to back the truck up a little bit the goal for both the filesystem implementation and your backup strategy is crash consistency. if the system crashes suddenly, due to a power outage, the filesystem driver is designed to ensure that all filesystem structures are consistent. filesystem data may be corrupted -- un-flushed data pages will be absent or inconsistent inside files -- but the filesystem itself will make sense and be parse-able by fsck or the driver. partly this is ensured by journalling the filesystem metadata. metadata is flushed to disk constantly, whether you want it to be or not. furthermore the kernel provides barriers for use by fs drivers so that nothing critical can ever be written without a flush, even by accident. when you start your dd backup at t+0, metadata updates will continue in the background, whether you like it or not. you may read a given directory structure at t+30, and then read it again at t+60, because in the background it was copied from one fs extent to another. the filesystem's anti-corruption strategy actively works against you to corrupt your backup. older filesystems did not do this, because it was dangerous. modern filesystems assume that all backups are taken from crash-consistent images. they make this assumption because it is, in the years 1995 and later, much easier to take a snapshot than it was in, say, 1985 do not pretend it is 1985 take an lvm snapshot and image that, not the disk. or do it offline, from a livecd. Notorious b.s.d. fucked around with this message at 19:30 on Jan 13, 2018 |
![]() |
|
nbsd's right here. you at least have to mount stuff ro if you want to dd it
|
![]() |
|
echi just get a Synology DS218+ (or 418), shove some 10TB drives in there, format as BTRFS, put Plex on it and call it a day.
|
![]() |
|
Tankakern posted:nbsd's right here. you at least have to mount stuff ro if you want to dd it re-mounting ro will not alter kernel processes, like the fs driver. if you are not going to take an lvm snapshot, the fs needs to be fully un-mounted i really do not get the resistance to lvm snapshotting. this technology is 20 years old on linux, 30 years old on legacy unix. it is just The Way Things Are Done, and always have been.
|
![]() |
|
also just in general i find file-level backups to be much more useful than block-level, because it makes partial restores a lot easier your pro strat for home linux backup is lvm snapshot + bacula or tar or whatever to actually do the backup. (if you put a little extra effort in, you can have bacula automate the whole process: create snapshot, refresh backup, delete snapshot)
|
![]() |
|
In my case I can't lvm snapshot because the partitions are not lvm.
|
![]() |
|
Colonel Taint posted:In my case I can't lvm snapshot because the partitions are not lvm. ![]()
|
![]() |
|
Colonel Taint posted:In my case I can't lvm snapshot because the partitions are not lvm. yeah you hosed that up real bad. you will never have reliable backups without lvm. i would fix that. the best part is you had to go out of your way to gently caress it up -- red hat and ubuntu both default to lvm for obvious reasons
|
![]() |
|
double plus fun: lvm is also the subsystem that makes it possible to migrate a running system from one disk to another so you can't even live-convert your system because you hosed it up lol
|
![]() |
|
Notorious b.s.d. posted:re-mounting ro will not alter kernel processes, like the fs driver. if you are not going to take an lvm snapshot, the fs needs to be fully un-mounted don't just make things up, dd from ro is fine. the issue people have with lvm is that it's overengineered and convoluted as gently caress if you just use it on a laptop/desktop harddrive. first PVs, then LVs, and _then_ the loving fs screw that
|
![]() |
|
but why not just use rsync or tar for backup, tar has finally catched up and is able to back up xattrs now too, making backups usable. bind-mount root to somewhere and backup from there so you dont have to exclude 100 dirs
|
![]() |
|
Yeah it's an arch linux install so setting up the lvm stack and making it bootable is a bunch of extra steps. I hadn't thought of bind-mounting the fs for tar but that does make it a lot easier so maybe I'll just do that.
|
![]() |
|
lvm is bad and lvm snapshots are the worst.
|
![]() |
|
just use windows which is easy to snapshot for reliable backups
|
![]() |
|
it’s as painful to setup as lvm in some respects but zfs beats the pants off of lvm after that. snapshots and backups become _too_ easy. btrfs can’t seem to get its poo poo together and the zfs licensing stuff is so stupid. supposedly there is some internal traction to relicense it but lol oracle. bcachefs seems fun with how you can tier storage but it’s super alpha.
|
![]() |
|
bcachefs, like zfs on linux, is a one man show, and i wouldn't touch it with a 10 feet pole btrfs owns, if you can bother to janitor your fs (like if you think zfs is an alternative), then really, go for btrfs. the stuff people mention like it doesn't handle going full or it eating data etc is isses from years ago, it's really production ready now. facebook has the devs for it now re: tar, remember to put on the args "--xattrs --xattrs-include=\*" to have ping work after restoring from backups (caps and selinux metadata and such is all xattrs)
|
![]() |
|
didn't redhat/centos remove btrfs or something
|
![]() |
|
yeah, to make their own fs management system replacing lvm suse still is all in on btrfs
|
![]() |
friend's mac keeps only one window active at the time, wtf? for example, opening a ppt attachment in outlook will cause all windows other than ppt minimise
|
|
![]() |
|
possibly the dock got set to single-application mode? http://osxdaily.com/2010/06/07/enable-single-application-mode-in-mac-os-x/
|
![]() |
|
![]()
|
# ? Jun 9, 2024 08:16 |
Colonel Taint posted:possibly the dock got set to single-application mode? http://osxdaily.com/2010/06/07/enable-single-application-mode-in-mac-os-x/ that is what ive thought of but they are an extremely non-technical user who is unlikely to have ever gone to terminal for anything
|
|
![]() |