00:33:44 ngortheone: sure, there's an argument to be made that there are degrees to reliability of protocols, but my gripe with email stems from the fact that on more than a handful of occations over the years when critical infrastructure has gone down, the notifications that were supposed to go out over email didn't. 00:36:52 debrup: sure. I was only poiting out that zabbix might also use email to alert, which weakens the case you were making for zabbix. What is your preferred alternative for sending alerts? 00:37:49 SMSes for everyone🎉 00:46:45 ngortheone: net-mgmt/alertmanager and www/rubygen-pagerduty generating notifications to my phone 00:47:21 If I still had a local SMS gateway, I might've used that. 00:52:24 ngortheone: https://pushover.net/ 00:52:26 Title: Pushover: Simple Notifications for Android, iPhone, iPad, and Desktop 00:52:54 Right, that's the alternative to paperduty. 00:53:00 I was trying to recall the name of it 00:54:26 don't like, 1 in 50 SMS messages just get lost, in the US? 00:54:38 I'm not in the US, though. 00:54:44 Email is more reliable than SMS, just on a general basis (i.e. no critical structure failures) 00:54:49 * xtile nods 00:54:57 Also, SMS has delivery and read-notification support. 00:55:02 I hate phones, I want to go phone-free this year 00:55:07 it has read notification? 00:55:11 isn't that RCS rather than SMS? 00:55:33 RCS definitely is an improvement over SMS, for sure 00:56:10 You're right, there's only delivery notifications - but if I'd setup something with my own SMS gateway, I'd retry if delivery confirmation didn't happen within a very short amount of time. :P 00:56:25 * xtile nods 00:57:11 I think what landed me on paperduty rather than pushover is that paperduty was in ports when I set it up. 00:59:08 pushover is just a curl call 01:00:25 Sure, and prometheus can do both pagerduty and pushover - but I needed something that could notify me in case prometheus decided to tip over. 01:06:28 a second prometheus 01:07:25 debdrup, pager duty? 01:08:05 debdrup, Nevermind; missed that you had later changed from "paPerduty" 01:09:57 someone knocks on the door with a card, on it written "httpd on thor is critical. [ ] ack [ ] snooze [ ] ignore" 01:19:38 a second prometheus 01:19:41 someone at work did this 01:20:12 two prometheus instances, one monitors everything and also the second prometheus instance; the second prometheus instance only monitors the first prometheus instance :P 07:04:55 so... I have a zpool. there's a disk I want to do some manual dinkery with, like using dd to write over a bad sector, hoping to clear the SMART Current_Pending_Sector... 07:05:35 can I get away with offlining that specific drive before doing this, or will exporting the zpool 'release' the drive so I can dd to it? or what? 07:05:49 I figure I'll need to resilver tit when I'm done. 07:06:11 there's 8 512 byte bad sectors (1 4K sector) 07:07:22 the drive is a 16TB WDC Gold, that came up with this single error 07:09:11 I'm hoping that if I can write 0s over the 'bad sectors', it will zero the Current_Pending_Sector 07:10:52 its a 512 logical 4096 physifcal sector drive, and the error is just the one 4096 physical sector, that shows up as 8 logical sectors. 07:14:05 so,, again, right now, I can't use dd or whatever to write on /dev/adao, but I can read from it. does if I offline the disk from the pool, will that be allowed? or do I have to export the whole pool? 07:15:30 You con put the disk in offline; then try writing over; do not know if ZFS would see that as the disk being online again. Also, so replace the disk while you tinker with th offending one 07:15:31 oh, if it matters, this is 13.1-RELEASE, in the form of TrueNAS 07:15:48 s/so replace/do replace/ 07:15:51 # uname -a 07:15:51 FreeBSD hognas.hogranch.com 13.1-RELEASE-p2 FreeBSD 13.1-RELEASE-p2 n245412-484f039b1d0 TRUENAS amd64 07:16:26 parv yeah, thats my intent, to resilver this drive after 'fixing' it. 07:16:49 Better would be put the disk offline, remove it, & tinker on a different 'puter 07:17:24 now, this is an entirely different issue, but I've been resilvering 4 drives from a raidz vdev, swapping 8TB drives with 16tb.... 07:17:58 What kind of RAIDZ is that -Z2? 07:18:05 first one took ~ 30 hours. 2nd one took 5+ days. 3rd one took 30ish hours. and this last 4th one is taking 5ish days. 07:18:08 *WTF?!? 07:18:34 its raidz cuz I only have 4 drives, its a HP Microserver N40L I deployed 10 years ago. 07:18:56 ( Wow, -Z1 ) 07:19:32 at my last $job (I retired almost 5 years ago), my preference was mirrors... lots and lots of mirros. 07:20:31 but with a 4 drive Microserver, thats not really an option, and this use case is basically archval bulk storage. 07:20:55 I do frequent scrubs 07:23:52 but, this last drive swap/resilver is saying its going to take 3 days 4+ hours more. and we're already a day+ into this resilver 07:29:28 Plan a picnic? Cook a turkey dinner? 07:29:45 hah. 07:31:02 i figure once this 4th disk resiover is finished, I'll export te whole pool and see if it lets me use dd to write directly on the 4K block thats erroring, then see if that clears the smart errors 07:31:16 if it doesn't, I'll triy using dd to zero the whole drive 07:31:39 [gawd, zeroing 16TB is going to take over a day at best] 07:32:54 or maybe I really should jsut RMA this disk to WDC 07:33:02 I bought it direct fromt hem 07:33:35 16TB drives for $250 each.. price on the gold drives has gone way up 07:33:49 ( was going to mention that; started a "dd if=/dev/zero" on a 4|6 TB disk from a old ZFS pool, but after ~10 minutes, gave up; clearing first 10 MB was enough for the purpose) 07:35:03 yeah, first few is fine for just making the drive 'clean', but that won't clear a SMART Current_Pending_Sector 07:35:24 Frankly I would go with RMA option 07:35:52 sigh, i hate having to deal with that but yeah, youre' probably right 07:36:08 a 'good' drive should NOT have bad sectors on the first use 07:36:31 even if its only 1 4k physical bad sector that mapps to 8 512 logicals. 07:36:36 Please buy 1 or 2 extra disks to have on hand, when next disk would go bad 07:37:08 my CFO (wife) probably would not allow that. 07:37:39 this 21-er-42TB NAS is mostly a media server for her brit TV 07:38:24 but I also use it for backups and stuff, its been running rock solid reliable for 10 years now. 07:38:40 which yes, I know, means I'm on borrowed time 07:38:42 Ok. Seems that downtown of days is tolerable 07:39:14 Damnit s/downtown/downtime/ 07:41:48 Why do you prefer WD Gold over Red Pro NAS ones? 07:46:17 um, they were on sale for $250 direct from WDC a couple weeks ago 07:46:48 and the golds are suposedly enterprise RAID grade, as opposed to 'red' home NAS grade 07:47:47 thats $250 for 16TB drives. 07:47:52 seemed like a really good deal 07:48:11 and direct from WD< not some reseller or grey market crap 07:48:37 my long term experience is, the less drives get handled, the more reliable they will be. 07:50:53 I used to get 20 drive cases direct from a major drive maker adn those drives were WAY more reliable than the brand had a reputation for. 07:54:09 All the WD Reds are not of the same color. Sure, avoid plain "Red" for NAS for using SMR tech. "Red Plus" & "Red Pro" use CMR. Do you know what is used for Gold ones? 07:54:30 https://www.tomshardware.com/news/wd-moves-to-settle-smr-hdd-false-advertising-class-action-lawsuit 07:54:31 Title: Western Digital to Settle SMR HDD False Advertising Class Action Lawsuit | Tom's Hardware 07:56:25 i've just configured a wireguard client on my 13.1 installation using the "experimental" kernel module. so far it works great. is there any timeline for merging into FreeBSD? the last commits to donenfeld's repo are from september, all i could find on the web is that an inclusion in 13.1 was planned 07:56:51 yeah, but I really hope these 16TB Golds, which are helium, aren't SMR 07:57:07 I knew aobut that whole red SMR shit, and wow, that was just stupid. 07:57:54 How does ZFS fare on SMR? 07:58:06 After all the COW does probably trigger a lot of rewrites too 07:58:31 very poorly, SMR tends to shit a brick during any sort of vdev rebuild 07:59:16 marketing around SMR has been borderline criminal 07:59:31 But I mean without zraid? 07:59:48 raidz, sorry... I'm thinking of migrating my ext4 nas to zfs (based on freebsd) now 08:00:08 I don't use raidz on it but instead snapraid on a separate machine 08:00:14 IMHO, SMR drive swould e just fine as a mag tape replacement device. 08:00:31 sequentially written, read randomly 08:00:49 That's how I back up now! I tar to a bunch of old drives in multi-drive format 08:00:49 but i like the spooling sounds of my lto3 :D 08:01:01 :-] 08:02:31 I cannot hear tape moving around over the noise of damn noisy fans going on always all the time (at work) 08:05:26 last tape drive I had at home was a DDS, I think it was 4gb DDS. I had LTO3 stuff at $job. 08:06:05 I've got an old sas-based lto3 and a fibre-channel based lto4 in my, eh, "homelab" 08:06:15 I do remember when 4GB was plenty of backup, HAH HAH HAH 08:06:24 [my home NAS is 42TB now] 08:07:00 i actually use them for backups, but just the family fotos and keepass files. not the whole NAS 08:07:25 my photos are backed up in the cloud 08:07:31 Work is using LTO-7 currently, ~6TB uncompressed; plans to move to larger capacity ones in near future 08:08:31 yes, if I lsoe the originals, I will only have the cloud versions, which aren't the full editable sources, but I havent' messed with lightroom in some years 08:08:35 so i don't care 08:09:56 i never figured out why i can't keep the lto drives at full speed when running the backups. according to the specs, the drive lto3 drive should max out at 80mb/s. the disks can definitely deliver that. but the drive spins up and down from time to time, i'm assuming it's a buffer thing 08:10:06 anyways, if I 'offline' a zfs drive, dd should be able to write to it? thats what I came here to ask 08:10:19 i tried plugging mbuffer in there, fine-tuning star parameters and whatnot 08:10:47 yet the drive always pauses and spins back up every 30 seconds or something 08:11:25 i heard it's bad for tape life, so i want to avoid it but never could 08:12:33 peerce, Yeah, remember to export the pool before tinkering 08:13:55 Wait! I thought peerce had changed to do RMA ... oh well 08:17:39 My home nas is > 100TB and tapes are probably just not feasible anymore 08:18:29 I use old drives now (to back up the most important 30% of it) 08:24:10 the latest lto9 has a raw capacity of 18TB per tape, but it's very expensive. tape only makes sense if you buy it used on the cheap or if you're a big enterprise with enough data to fill so many tapes that the lower per-tb price of the indivudal tapes offset the high system price 09:06:26 Yeah exactly, this is why I use my old drives as tapes 09:06:43 The LTO drives are so expensive 09:06:54 And also too noisy for a home environment IMO 09:07:46 The drawback is that my backup pool is always too small as I always replace my drives with bigger ones 09:10:15 wildeboskat, Quite obvious that you ought to stop buying larger capacity disks ;-J 09:10:29 But then I don't need to replace any :) 09:10:44 Well, yeah 10:11:37 Like other I always need to upgrade my backuo sysème. I need to change 3TB disks of my NAS with may be 8TB, to be sure to run my jail and store my backup. 10:11:56 *system 10:13:07 there a safe way to lower the ARC for ZFS? Running into an issue where the machine is running out of RAM a lot and going into swap space 10:15:10 dedup is off, but compression is on.. only get 1.04x from that so might turn compression back off 10:29:34 Grabunhold: I find LTO-6 a good middle ground because you can buy decent LTO-5 tape for cheap and format it to 6 without any issues 10:30:50 2.5 TB uncompressed is pretty good 10:32:03 Remilia: well, my lto4 is okay for the most important of data like the family fotos and stuff. of course i'd like more, but even the used tape drives are still pretty expensive... 10:32:15 true 10:32:31 and you need SAS :\ 10:33:23 LTO tapes.. wow I haven't used those in about 17 years 10:33:35 people still use tapes to back up files these days? o_O 10:33:53 considering it is the best option for long term storage, yes 10:35:55 pertho: do you have any better suggestions when it comes to density/price/longevity? 10:36:11 I remember once putting a LTO tape in a demagnetizer.. and the demagnetizer couldn't even get through to scramble the tape.. checked the tape afterwards and all the data was still there. I think you'd need some powerful electromagnet to kill an LTO tape :D 10:36:39 well, i got the tape drive for free at work because they wanted to toss it. came with a bunch of tapes, too. so why not dump the family fotos, keepassxc database and stuff on there every month. makes a decent backup and i can even easily carry the tapes to the office for "off-site" backup 10:37:14 Remilia: sorry, no better suggestions aside from perhaps having both onsite and offsite backups if possible 10:37:15 I want an internal LTO drive but the price is staggering 10:37:38 you can never have too many backups 10:39:12 i've got an internal fibre-channel LTO drive. had to take a dremel to my case because apparently double-height 5.25" drives aren't a thing in consumer computercases. and the cheap fibre channel controller from ebay has no internal connectors, so i'm re-routing the fiber back into the case via a missing PCI bracket :D 10:52:57 pertho, You could adjust vfs.zfs.{arc.max,arc_free_target} MIBs; Allan J metions *.arc_free_target MIB in https://youtu.be/L_kBM_RHa6g?t=1533 10:52:58 Title: ELI5: ZFS Caching (2019) - YouTube 10:56:35 Set "*arc_free_traget" to keep that amount free. Note that size is specified as number of pages ; "hw.pagesize" tells you current byte/page 11:05:01 thanks! 11:18:57 mason, I came across your post https://wiki.freebsd.org/MasonLoringBliss/JailsEpair (I'm assuming that is you) - I replaced lines 5 and 6 in the jail.conf example with the dhclient line and it's working nicely 11:18:58 Title: MasonLoringBliss/JailsEpair - FreeBSD Wiki 11:38:58 pertho: not sure if you got answer but her goes... from iwm(4) man page: Currently, iwm only supports 802.11b and 802.11g modes. It will not as sociate to access points that are configured to operate only in 802.11n or 802.11ac modes. 12:12:25 ahh thought as much, thanks 12:57:07 bsd.to has an invalid certificate 12:59:04 is freetype compiled with png support in FreeBSD? 13:00:00 freetype2 13:00:30 pertho: Appears to be. 13:00:40 I think rinfo is a default pkg alias? 13:00:51 pkg rinfo -f -r FreeBSD freetype2 13:01:15 er no it's not. 13:01:17 /usr/ports/print/freetype2/Makefile has "OPTIONS_DEFAULT= BROTLI LCD_RENDERING LONG_PCF_NAMES PNG V40" 13:01:46 On my default pkg: PNG : on 13:01:52 hmm 13:02:05 Shared Libs required: libpng16.so.16 13:02:42 pkg search -S "name" -e -f -r FreeBSD freetype2 I have "rinfo = search -S "name" -e" in pkg.conf 13:02:49 I don;t see rinfo as a pkg option? 13:03:16 pertho: Yeah, I forgot it's a custom alias. 13:03:34 The second command is the same. 13:04:25 If you're dealing with mutliple repos then you can drop the -r FreeBSD part or change it to the repo of youre primary install. 13:05:07 hmm ok I see same thing: Shared Libs required: libpng16.so.16 13:05:28 vkarlsen, re bsd.to, maybe best to email koobs. 13:05:39 trying to resolve an issue with emojis from noto color emojis not showing properly in an app (they also don't show up in "gucharmap" either) 13:07:07 Good day everyone 13:07:07 I have a peculiar problem on FreeBSD 13.1 that is very strange: when i use frr + full route import, not only does zebra use 280% cpu, bgpd 100% cpu, but it seems to also cause other processes such as charon, ntpd and dnsmasq also to consume 100% cpu 13:07:27 I have raised kern.ipc.maxsockbuf to 16777216, but aside from that the rest of the system is rather stock from opnsense 13:07:47 V_PauAmma_V: Ok, done 13:08:03 I was wondering if anyone has a hint on what might be wrong? I checked the usual search terms for perforrmance tuning, but did not spot anything useful so far 13:09:55 skered: Thanks, that's a neat alias. I put it in my own pkg.conf o/ 14:14:00 telmich: out of curiocity, what CPU state is it in? 14:50:02 cation: Yeah, there are a couple things I changed to accomodate bugs that seem to have gone away. I want to go through to remove this, marking older mitigations and options. Thank you for the feedback! 14:50:26 s/this/these/ 14:51:37 sure, and thanks a lot for writing it :) helped me a lot 14:57:27 cation: One thing I want to explore is actually not having any logical (just physical) network config in the jail config, and having the system itself do that work. I believe it was meena that noted that interface names can be set to something predictable, so that ought to do the trick. 14:57:51 If I test this out and have success I'll note it alongside the existing methods. 15:01:40 oh, that would be neat 15:17:46 debdrup: You mean he cpu state of zebra? 15:18:11 debdrup: `15792 root 9 103 0 1400M 615M CPU11 11 339.2H 277.97% zebra` 15:18:33 And for bgpd: 29426 root 4 103 0 68G 50G CPU9 9 121.5H 100.19% bgpd 16:40:09 anyone in here been using Fly.IO platform? and with that flyctl? 17:14:27 i tried fly.io 17:19:41 telmich: I'm honestly not sure 17:56:46 looks like Fly.io is the backend for supabase deployment. very cool 18:18:35 rtprio: have you used flyctl in freebsd or what tools did you use to deploy to the platform? 18:59:38 xD grub-bhyve segfaults for me when trying to boot a CentOS image. A great start to the day lololol 19:14:36 Are there any plans to make /bin/sh as the default root shell in 13.x? 19:14:55 That'd break POLA. 19:17:21 Alright 19:19:10 That was my understanding too. But could change in some future 13 version, for the better despite POLA AFAIAC 20:02:07 parv: you're gonna have to use more words to explain that sentence, because it doesn't make an y sense. 20:05:32 debdrup, "But could change in some future 13 version, for the better despite POLA AFAIAC", rephrased: If some future 13.x version would make "/bin/sh" as the default root shell in spite of POLA, I would welcome that for I do prefer "/bin/sh". AFAIAC "/bin/t?csh" could disappear for ever 20:06:57 parv: the proposal to change roots shell doesn't include an MFC notice, and it's even explicitly mentioned in UPDATING - so why would you assume it changes in any minor version of 13? 20:07:14 debdrup, I did not assume. I am wishing 20:07:59 Also, your shell preference has absolutely nothing to do with what should be in the base system, and the proposal - if memory serves - explicitly confirms that the change was _not_ supposed to be about removing (t)csh. 20:08:51 It's been a long time since I've seen a user be this arrogant, and I'd appreciate it if you take it _ANYWHERE_ else. 20:08:59 Yeah, I know /bin/t?csh would not be going anywhere. Aagin, wishing 20:09:17 What does it matter to you that tcsh is in base, if you don't use it? 20:10:48 All the documentation is Bourne-compatible, there are (as far as I know) no scripts that depend on the C shell in the base system, and there's a significant number of committers who do use (t)csh. 20:14:02 I like tcsh, though I believe we are not too many liking it lol 20:14:52 I'm fairly sure there's more than people seem to assume. 20:19:06 Am I correct, that FreeBSD kernel moves swapped out memory back to RAM on demand? 20:22:04 mrtnt: I think you're confusing terms a bit; things that can be swapped out of memory include clean pages, clean anonymous pages (dirty anonymous pages are cleaned during the process of writing them), sysv ipc, as well as filesystem data such as files and metadata. 20:22:31 If something is swapped to disk and it's referenced, then yes it will be brought back into memory - that's kinda the point of paging. ;) 20:24:39 It's a fair bit more complicated because all of that doesn't account for the unified buffer cache that FreeBSD has, the active, inactive and laundry queues of the pagedaemon, and it also doesn't take into account ZFS' ARC. 20:25:10 And then there's stuff that, like ZFS ARC, is wired and therefore will never be swapped. 20:30:49 debdrup: ok. So even if something is swapped out to a swap partition or swap file and there is a need to access this swapped out data, then it will always be brought back to RAM. 20:35:39 mrtnt: yes, although since you mention swap file, I think it's important to mention that swap files generally speaking aren't recommended. 20:37:12 debdrup: ok, understood. Thanks! 20:42:03 * Ellenor whistles 20:58:53 re: swap, what's recommended for small RAM VPS installs (eg: 1GB RAM)? 21:45:31 ISTR seeing "2x RAM if < 4GB RAM" 22:06:48 That's a good rule of thumb, yes. And > 4GB, swap should equal the amount of memory. 23:17:12 spork_css_: it depends on the ram size. if there's 32GB ram i make 4 swap partitions of 4gb each. Then if stuff starts swapping and stays swapped, one can swapoff that partition then swapon to force whats swapped out back into ram. This approach makes less sense much after 32GB ram 23:18:24 if it's all just one big swap partition then you can't do that unless the amount in swap is less than the amount of free ram 23:22:30 ^^ that should read 4 swap partitions of 8gb each 23:58:26 f451: that's a lot of swap