03:03:10 Hi, has anyone ever encountered an issue where typing `ls` into a shell yields output in which one of the directories is misspelled? For example, for 3 consecutive times, when I typed `ls`, the name of the `shared` directory was spelled `shaked`, and then it went back to normal. 03:03:34 "rtyler, I am just a newbie to..." <- NB not a new dataset within the same pool. 03:06:56 torbo: Maybe a problem with terminal emulator and $TERM variable? 03:08:16 With `st`, sometimes characters get "stuck" on weechat, so it's like it doesn't fully draw lines. Telling weechat to do a full redraw usually fixes it. So maybe something similar is going on. 03:09:08 (With `rxvt-unicode` this problem is way more uncommon for me.) 03:10:27 Thanks, shiroyasha, this was in an Emacs shell mode, and on MacOS, to be specific, so not FreeBSD-related (though MacOS is based on FreeBSD) -- I do run FreeBSD on my desktop. It's just never happened before, so it was particularly surprising. I did take a screenshot and definitely the `k` character was drawn instead of the `r` character. The characters look very different in this font. 03:14:13 torbo: Just to clarify, if it's the same problem I describe, then it would be that there was a "k" in that row+column before you ran `ls`, but after running `ls` that row+column was left unchanged, keeping the "k" from before (unchanged) instead of replacing it an "r" as expected. 03:14:39 Emphasis on "if". It might be something different. 03:16:39 Yeah, you're right, there was a k in that very column in the previous row. Maybe the character re-drawing logic is glitchy and a pointer wasn't de-referenced properly (or something). 03:16:55 I don't know anything about Emacs, but if that "shell mode" is a terminal emulator, I would try setting `TERM` environment variable to "screen-256color" or "xterm-256color" and see if that fixes the issue. 03:19:29 Thanks, shiroyasha. The `TERM` environment variable is already set to xterm-256color. It's just odd that I had never encountered this issue before in over a decade of Emacs use. It happened on my work computer and I worried that it may have been compromised, especially since the word is s_haked_. 06:33:41 you could also try setting it to vt100 which is the simplest thing supported by everything, and checking with that 06:37:39 torbo: a little. 06:37:40 > … (though MacOS is based on FreeBSD) … 06:39:29 historical, condensed. 06:39:30 Title: grahamperrin comments on Practical reason to choose FOSS GNU/Linux over FreeBSD: 06:40:56 A few months ago I found a very neat summary of the relationships between FreeBSD and Mac OS X, I couldn't find it when I wrote that. I should have bookmarked it. 08:13:23 sysutils/ztop <3 11:26:34 I need to debug how an aliased hostname gets resolved. I can ping my "alias-hostname", but I do not see anything related to this host in places like /etc/hosts. Any tool or "verbosity" switch saying the way this hostname gets resolved? 11:37:09 See truss(1). 11:39:32 Or (since you want to trace resolving and it's likely DNS if not /etc/hosts), use "options debug". See resolver(5). 12:16:36 How to install the video drivers in a qemu freebsd virtual machine? 12:41:13 sukamu: i installed from source a patched version of xf86-video-qxl 12:57:50 hi, any idea how I get the source code for FreeBSD 13.1 ? 12:58:53 svnlite seems not to be on 13.1 12:59:46 When you install it, bsdinstall gives you an option to install the source. Otherwise, there's a way to fetch it and untar it later... hold on, lemme get the URL. 13:01:52 svnlite checkout svn://svn.freebsd.org/base/releng/13.1/ /usr/src 13:02:01 if I install svn package, will this still work? 13:04:04 I think this is the url for src: http://ftp.freebsd.org/pub/FreeBSD/releases/i386/13.1-RELEASE/src.txz 13:05:03 (I will fetch this and install) 13:06:36 is the svn mirror still maintained? the source migrated to git a while ago 13:07:25 It is, although I'd use the one for your architecture, eg https://download.freebsd.org/ftp/releases/amd64/13.1-RELEASE/src.txz or https://download.freebsd.org/ftp/releases/powerpc/13.1-RELEASE/src.txz . 13:08:49 It is, IIRC. Just not updated, but still needed for history of files that were deleted or renamed before the git conversion. 13:41:00 this is more a macos queation, but i think freebsd uses a similar route tool 13:41:23 trying to understand what command creates this route -- fc00:2:: f6.d4.88.d6.75.64 UHL lo0 13:41:57 i don't know what it means that there is no /64 there or any subnet info... and also not sure where the mac address is coming from 14:13:28 whats happening with the new libXft with bgra support? Is it going to make it into 12.3 anytime soon? 14:20:35 V_PauAmma_V: thank you! (right now I am using i386 for a nanobsd image for some embeded project) 14:21:31 paulf: that I was curious, if svn is still maintained 15:06:06 paulf: meandrain: only for12.x 15:06:40 stable/12 and releng/12.x are imported only 15:07:44 so for 13 instead of svn, I shoul use git? 15:09:17 yes 15:34:15 Hi , if someone cana have a look : https://twitter.com/miconof/status/1575507579584888832?ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed%7Ctwterm%5E1575507579584888832%7Ctwgr%5E0c1e6acc53252ee29b1f18908658cee82a3a2b14%7Ctwcon%5Es1_c10&ref_url=https%3A%2F%2Fpublish.twitter.com%2F%3Fquery%3Dhttps3A2F2Ftwitter.com2Fmiconof2Fstatus2F1575507579584888832widget%3DTweet 15:34:16 Title: Michel L on Twitter: "@OpenZFS encounter a bug with #ZFS under #FreeBSD 13.1 on a #NFS export trying to access to 'vol/.zfs/snapshot' form NFS give a 'Stale file handle' and then if you try to delete the #snapshot -> zfs hangs : see : https://t.co/LMszWfvWj5" / Twitter 16:25:24 I am trying to compile freebsd 13.1 i386 and I get this error (compiling kernel): https://pastebin.mozilla.org/ScWZ5ieU 16:25:25 Title: Mozilla Community Pastebin/ScWZ5ieU (Bash) 16:25:27 any hint ? 16:26:18 rwp: media sector size is definitely 512, I wonder (aloud) if that's something that can be changed live without re-imaging. 16:28:52 https://pastebin.mozilla.org/SdnYMDCR 16:28:53 Title: Mozilla Community Pastebin/SdnYMDCR (Bash) 16:32:26 I found the problem, the file /usr/src/sys/contrib/zlib/deflate.c is borked somehow. 16:32:57 meandrain: here that line looks like "z_streamp strm;" 16:33:03 yep, was just writing that 17:03:23 rwp: the verbose status outputs fwiw https://bsd.to/vZuP/raw 17:03:24 Title: vZuP 17:04:33 rtyler, If all of the drives in your storage array are 512 byte physical sectors then that's fine. But if any of the drives are newer 4k AF drives then that's definitely *a* performance problem! 17:04:58 Your status shows resilvering in operation now. So that will also slow things up until that completes. That's normal. 17:05:29 yeah, that I just started. I had taken a disk offline just to play a little while ago to see if any performance changed 17:05:48 what's interesting to me is how freaking slow that resilvering is 17:06:42 everything in the array is 512k sectors fwiw 17:06:53 er, just 512 :) 17:10:02 I like examples. Here is one from one of my drives. https://bsd.to/BUOR/raw 17:10:03 Title: BUOR 17:11:11 <___used> Hi. I am new here. Used *BSD 10 years ago, not since then. Looking at VPS offers, low end, I see 1GB RAM claiming supports it. Is this realistic? Would fbsd run in 1GB RAM with say lighty or a stripped version of apache and a db? 17:11:53 And here is an edited paste with an actual 512 byte sector drive example added. https://bsd.to/v77d/raw 17:11:54 Title: v77d 17:12:27 https://bsd.to/4OQe/raw 17:12:28 Title: 4OQe 17:13:08 rtyler, That is definitely a 4k AF advanced format drive and if ashift defaulted to 9 for you then that is definitely something that should be fixed. :-( 17:13:31 Hello ___used. First, I personally think it should work but am rather a newbie myself and so cannot say. 17:13:59 I think it should. But the best answer would be to try it. Since it is a VPS it is easy to try and test. 17:14:55 <___used> Experience: the best teacher, you get to find what works after trying out everything else? :-/ 17:16:00 rtyler, What does "sysctl vfs.zfs.min_auto_ashift" say for you? I think it must say 9 (as in 2^9 = 512) but desired is 12 (2^12 = 4096). 17:16:02 it definitely should, depending of course what apache is running 17:16:17 rwp: it says 12 17:16:30 12 is good. 17:16:47 Was this array created some time ago when the default was 9? 17:16:56 <___used> I said stripped down for a reason... Also, anyone running Asterisk or similar, also limited/stripped down on fbsd? I'm interested in RSS footprint and such. RAM. 17:17:24 I want to double check and verify the array ashift size. Best to know for sure before becoming too concerned. 17:17:35 rwp: it's been around for about 2-3 years I think, this was a 12-RELEASE box originally, and I'm pretty sure I built it in 2020 17:21:12 I don't know when the defaults were updated from 9 to 12. I am having a mental stall trying to remember how to ask the array what ashift value has been set. 17:23:34 * rtyler ponders 17:25:05 `zfs get all zroot` doesn't show me anything interesting 17:27:18 "zpool get ashift zroot" shows?? Perhaps "zroot ashift 0 default" ?? 17:30:16 Example from this end. This was also created on 12.x (I forget which) https://bsd.to/mH6B/raw 17:30:17 Title: mH6B 17:31:48 ___used, For memory efficiency I am an advocate for Nginx over Apache. It has a hugely smaller memory footprint. (If I can say "hugely smaller"!) 17:32:19 <___used> Yes, or lighty. Or even thttpd for non SSL. 17:35:00 rwp: yeah, we're looking pretty identical 17:35:20 https://bsd.to/cb9x/raw 17:35:21 Title: cb9x 17:36:11 I think your array is actually using ashift of 12 as desired. But I don't know enough to know for sure. Sorry. Hopefully someone with real knowledge will know. 17:37:25 hah, thanks for the help nonetheless 17:37:47 maybe some time next year when this resilvering is complete I'll be able to rebuild the array xD 17:38:13 working at leass than 1MB/s is pretty gnarly 17:38:20 2.95G scanned at 804K/s, 2.09G issued at 569K/s, 2.21T total 17:39:52 Just fyi but here is the product manual for your drive https://www.seagate.com/www-content/product-content/barracuda-fam/barracuda-new/en-us/docs/100817550b.pdf 17:44:19 <___used> Something is seriously wrong for a SATA drive and such slow speeds. 17:44:31 <___used> Any io errors or scsi errors in logs? 17:44:34 I looked at the spec page because I was worried it was perhaps an SMR drive and those are really problematic for resilvering. 17:44:39 But that page never mentions shingle anywhere. So okay. 17:45:21 <___used> Does that mean it is not shingle type? ;) 17:45:47 <___used> Shingle type saves money for the maker. Was it cheap or an "AV" disk?... 17:46:20 <___used> DVRs are one/the(?) last bastion for spinning rust sales now. 17:46:36 <___used> And they use shingle mode disks of high capacity usually. 17:47:18 what? people buy nases all the time 17:47:59 <___used> Sure, but for consumer level low cost drives, shingle mode "AV" disks are the standard. Your average convenience shop DVR. 17:49:29 I literally dont care about that shit, but sure, I'll take yer word for it 17:49:36 <___used> np 17:49:47 shingled are cheaper for the consumer too 17:49:52 <___used> It's just a factoid to keep in mind when buying AV drives. 17:50:02 <___used> And using them for non AV things. 17:50:13 <___used> Random access being one of the other things. 17:50:20 It is true that SMR is cheaper and since 2020 problematic for the buyer using raid not knowing of the problems. 17:50:27 I can see them being used for AV purposes 17:50:43 SMR gets you more density per platter, so they end up being cheaper per TB 17:50:50 <___used> Random access write in particular. 17:50:55 Before 2019 SMR didn't really exist and so wasn't a concern when buying drives. But since 2020 it is now a big concern and easy to buy a drive "not suitable for purpose". 17:51:25 SMR Problem Reference: https://raid.wiki.kernel.org/index.php/Timeout_Mismatch 17:51:26 Title: Timeout Mismatch - Linux Raid Wiki 17:51:39 storage is hard for lots of people 17:51:40 Which also is a good reference for ERC SCT too. 17:51:52 lots of things to think about, lots of marketing bullshit 17:52:08 lots of things you'll never think about until you hit it 17:52:29 Western Digital's marketing department got caught mislabeling drives. They came clean. https://blog.westerndigital.com/wd-red-nas-drives/ 17:52:30 Title: On WD Red NAS Drives - Western Digital Corporate Blog 17:52:38 <___used> Also, pretty much no-one writes smr on disk specs. That I can tell. 17:53:11 Toshiba says they never screwed that up https://toshiba.semicon-storage.com/ap-en/company/news/news-topics/2020/04/storage-20200428-1.html 17:53:12 Title: Use of Shingled Magnetic Recording (SMR) technology in Toshiba Consumer Hard Drives. | Toshiba Electronic Devices & Storage Corporation | Asia-English 17:54:08 It's definitely a problem to be concerned about but it is possible to step carefully around the problems. Information is power in this case. 17:55:50 <___used> There was that story about someone buying a bunch of wd I think drives for NAS use and the second time he bought the same kit set they were different, with same model number. 17:56:11 yes, I've heard that story 17:56:22 scary crap 17:56:55 Yes. WDC got caught doing that. They later came clean and published a map so buyers could know which was which. That was in their Red line of drives. 17:57:08 i have a bunch of red 17:57:20 (at least 16 red drives here) 17:57:22 nacelle, Please read https://blog.westerndigital.com/wd-red-nas-drives/ then. 17:57:23 Title: On WD Red NAS Drives - Western Digital Corporate Blog 17:57:34 rwp: I'm not new :_) 17:57:41 already done :-) 17:57:49 Okay then! :-) 17:58:03 <___used> not an ad, typical listing of AV drives. https://www.cclonline.com/storage/hard-drives/Surveillance-AV-Hard-Drives/ 17:58:04 Title: Surveillance (AV) Hard Drives | CCL 17:58:27 FTR I have a bunch of WDC Red drives too. 17:58:28 <___used> No trace of SMR but you can bet they are. Large cap, slow speed, low price. 18:00:32 https://www.seagate.com/internal-hard-drives/cmr-smr-list/ 18:00:33 Title: CMR and SMR Hard Drives | Seagate US 18:01:13 those seem to be CMR 18:01:23 8tb isnt a lot nowadays 18:01:31 <___used> https://hdd.userbenchmark.com/ turn on '4k write' column using >> arrows link on right 18:01:32 Title: HDD UserBenchmarks - 1015 Hard Drives Compared 18:01:47 you can get 14tb used enterprise disks off the eebs for under $170 now 18:01:54 I think the WD Red HDDs are also SMR now... 18:01:59 not all 18:02:06 there are some that are and some that arent 18:02:13 WD Red enterprise are the new CMR hdds 18:02:34 * ___used remembers silly jokes about Helium filled disks needing to be taped to the desk to avoid them flying to the ceiling 18:02:43 and red plus disks too iirc 18:03:36 https://www.servethehome.com/wd-red-smr-vs-cmr-tested-avoid-red-smr/ 18:03:37 Title: WD Red SMR vs CMR Tested Avoid Red SMR - ServeTheHome 18:04:08 <___used> Surprizing to find 2018 disks (Barracuda 2TB) writing at only 0.79 MBps ... see my link above 18:04:48 <___used> 0.79 MBps is slow even by 2010 USB 2.0 connected dollar store usb stick 18:06:34 <___used> SSHD is very nice for random write speed ;) 18:07:37 do red drives have the 3 or 5 year warranty? 18:08:04 you have to look at each drive specifically to find that out 18:08:10 by in large 5 though 18:08:18 <___used> wow, great marketing 18:08:21 theres no generic "All WD Reds are... " 18:08:28 they've been around for over a decade 18:08:32 what do you expect? 18:08:43 <___used> A listing. Perhaps 3rd party. 18:08:49 they have that somewhere 18:08:51 i've just been buying wd black. 18:09:20 WD Blacks dont do TLER, which makes them iffy for raids 18:11:17 i've had them in the same zpool for 9 years :shrug: 18:12:29 And advantage of that era of drives is that they were excellent drives. 18:12:45 It's the new drives now that one must be cautious of since things have changed. :-( 18:13:08 <___used> More exactly: new drives with the /same/ model number as old. GRR 18:13:37 i haven't looked at storage in a long time; i'm due to rebuild the pool as it's half 512 and half 4096 sectors 18:18:06 i'm in the middle of a storage project brought on by a failing disk in one of my hand built nas boxes. 18:18:45 <___used> Is there a port of freebsd for synology etc nas boxes? 18:19:06 for all the hoopla about the wd red 6tb disks being not great, I've had a set of them that went without errors for 7 straight years, and now I've only one thats failing. (I have a bunch of REDs, this is just my 6tb set) 18:25:17 ___used: yeah, this performance isn't what I would expect from a SATA disk at all, maybe a USB 1.0 flash drive :P 18:29:07 maybe I left a "suck jumper" on the drives 18:30:19 <___used> There should be no jumpers on a SATA disk? 18:32:07 'twas a joke :) 18:32:50 <___used> I could not tell ;) 18:33:47 <___used> I wonder how WD's DMSMR "idle time available in SOHO use" ties in with Microsoft's prducts being seemingly designed to keep the HDD led hard on... 18:34:04 <___used> +o 18:36:36 Hi. Could someone post the default permissions for the /var/log folder? I changed it to exclude 'others' with "chmod o= /var/log", but now the polipos service won't start due to insufficient permissions. 18:41:15 755 root:wheel 18:41:34 pmnw: /etc/mtree/BSD.var.dist 18:41:50 i.e.: /set type=dir uname=root gname=wheel mode=0755 tags=package=runtime 18:44:50 any idea how to fix this build error (freebsd 13.1, i386): https://pastebin.mozilla.org/KrzSSViP 18:44:52 Title: Mozilla Community Pastebin/KrzSSViP (JavaScript) 18:45:05 <___used> The way to have more private logs is mkdir /var/log/moreprivacy and set perms on that if you have to 18:45:31 meandrain: i'd suggest to open a PR for it, if there isn't any already. 18:46:01 <___used> PR means pull request here?! 18:46:16 otis: neat - thank you 18:50:44 * rtyler ponders where to search next for this performance problem 18:50:52 honestly I feel like the solution is to just rebuild the machine entirely 18:50:59 <___used> rtyler: find your disk's real speed. 18:51:13 <___used> rtyler: f.ex. on the link I posted above 18:51:15 i will open a PR but first I want to be sure it's not something I am doing wrong 18:51:39 <___used> rtyler: random read and write speed 18:51:55 <___used> You can't make it go faster than the manufacturer provided. 18:56:07 running the naive benchmark from `diskinfo -t` 18:56:18 on a disk other than the one which is resilvering of course 18:57:41 <___used> The benchmark does not test random read write I think, only contiguous read or write? 18:58:03 yeah, doesn't do writes. 18:58:20 the only other tool I'm familiar with is bonnie++ 18:58:47 <___used> Even so, random read takes into account head positioning speed, which can slow down things a lot, as well as spin latency (rpm). 18:59:10 <___used> Remember screaming Barracudas and Fireballs? 15krpm? 18:59:54 cheetahs? 19:00:04 <___used> That too I think. 19:00:18 <___used> I remember the spinning up jet engine sound... 19:00:33 (I'm not sure that Barracuda or Fireballs ever came in 15k form fwiw) 19:00:47 <___used> Could be. It's like 20 years back? 19:00:54 I had some in a 4U rack back in the day, a whole array, loud AF 19:01:10 i've been in data centers full of them 19:02:03 couldnt really hear the drives over the fans 19:02:24 <___used> I had one in a normal PC case iirc. Without covers. 19:02:35 where the fans are of the "ear bleed" type 19:02:47 so let's pretend that the disks aren't trash and I need to just reimage and reconfigure the entire zpool, I'm assuming since this is a boot to ZFS environment, I'm going to have to send the datasets somewhere else for a period of time and then bring them back after the new install? 19:03:02 <___used> https://www.alphatechserve.com/all-categories/storage-devices.html fiber channel ones are still made 19:03:04 Title: Storage Devices - All Categories 19:03:27 <___used> The server fans are small and mean. Someone made a hoverboard using several. 19:03:48 made? 19:04:14 <___used> There is no choice but to spin fast in a 2in tall chassis to move a lot of air. 19:04:28 <___used> built/assembled. Tethered, not self powered. 19:05:01 i mean, the fiber channel thing - whats the point of that url? 19:05:07 it shows a reseller of old crap? 19:05:18 <___used> Just exemplifying 15k drives you can still buy. 19:05:47 you can buy anything on ebay ;-) 19:06:41 ___used, In FreeBSD PR == Problem Report reference https://docs.freebsd.org/en/articles/problem-reports/ 19:06:42 Title: Writing FreeBSD Problem Reports | FreeBSD Documentation Portal 19:07:27 <___used> ok, github has a de facto monopoly on PR... I was guessing there. 19:08:02 PR was Pull Request well before github 19:08:19 <___used> I think it was Public Relations before it was stolen ;) 19:08:27 PR has been Problem Report from like the last century. 19:08:57 Github created the pull request as far as I recall from history. Because git created "pull". 19:09:53 Before git no one pulled or cloned anything. Before it was always "checked out" and a few other names. YMMV! 19:10:07 <___used> yes 19:10:46 rwp: yes, it came with git, not, it did not come with github afaicr 19:10:57 (git came well before github) 19:11:30 its plausible my memory of the events are wrong, I concede this 19:13:26 https://en.wikipedia.org/wiki/Git "Pull requests are not a feature of git, but are commonly provided by git cloud services." But let's take this ot discussion to -social if we are just going to chat. :-) 19:13:27 Title: Git - Wikipedia 19:33:37 ah I thought I was in social 19:33:45 I see freebsd-irc and think thats the social, my bad 20:07:44 here's some disk performance details from a device: https://bsd.to/ks1d/raw not sure how to parse this offhand however 20:07:45 Title: ks1d 20:07:57 _seems_ slow 20:10:00 try -cit 20:19:48 uno momento 20:21:30 heh, I hear the disk fiddling away in there 20:28:59 https://bsd.to/gzC8/raw 20:29:00 Title: gzC8 20:29:36 dozens of IOPS, dozens! 20:42:38 debdrup: I don't entirely know what to expect here, but the transfer rates at least look good to me in terms of MB/s 20:42:46 heh, sectorsize: 1131643 ops in 3.000032 sec = 377210 IOPS 20:43:11 (it's emulated nvme, so not real numbers) 20:44:27 heh, yeah the IOPS aren't great xD 20:44:27 hi, I would like to test suspend/resume in my desktop (so we save some energy without having to power off) 20:44:30 what shall I do? 20:44:39 which commands suspend? then is it just pressing a key? 20:44:46 zzz 20:45:53 rtyler, I ran this on a VM not bare metal: https://bsd.to/SRUe/raw 20:45:54 Title: SRUe 20:49:25 uskerine, In case yuripv's response was perhaps taken to be a comment about sleeping, please look at "man zzz". 20:50:14 rwp: your numbers seem to indicate that these disks are just garbage 20:51:00 aside from throwing them into the ocean and replacing them, I'm not sure what can be done. I might take the machine down and poke to see if there's any flags in the BIOS to fiddle with 20:52:17 I haven't run diskinfo before. It's entirely safe on a live file system that I don't want to corrupt? 20:53:29 heh, I'm running it with reckless abandon on my live system XD 20:53:52 Also, did you paste in your "smartctl -i /dev/ada0" information? 20:54:06 yeah, I pasted that before, lemme repaste it because who knows where it is 20:54:48 smartctl: https://bsd.to/Xdb3/raw 20:54:49 Title: Xdb3 20:56:44 Another thing I spot there is this "SATA Version is: SATA 3.1, 6.0 Gb/s (current: 3.0 Gb/s)" 20:56:58 Not a deal breaker but that is a 2x difference in speed due to the SATA port. 20:57:25 The real killer is "Model Family: Seagate BarraCuda 3.5 (SMR)" SMR! OUCH!!! 20:57:42 anyone here using chrony on your FreeBSD servers / vms? 20:59:00 rtyler, Go read this https://www.servethehome.com/wd-red-smr-vs-cmr-tested-avoid-red-smr/ jump to page 2 scroll to the RAIDZ Resilver Time graphs. Have a good cry! 20:59:01 Title: WD Red SMR vs CMR Tested Avoid Red SMR - ServeTheHome 20:59:59 There a way with portsnap to tell it to ignore certain ports collections? 21:00:06 meena, I am using ntpd on mine. 21:00:25 rwp: as in the default one? 21:00:28 yeah, I don't know why anyone would run chrony on FreeBSD. 21:01:02 meena, Yes. The one in core. 21:01:59 https://man.freebsd.org/ntpd(8)#BUGS :P 21:02:00 Title: ntpd(8) 21:02:24 rwp: SMR is trash, got it :D. Briefly reviewing what SMR is and how it works, I suppose it's not surprising that as data has increased on this pool that the performance pain has become much more noticable. I'm not sure if I've got anywhere to go other than replacing it all :/ 21:02:55 tbh, the main reason why i would consider using something other than ntpd is why I would consider using something other than sendmail 21:03:01 meena, I read that as more of a social comment than as a real bug. It might be heavy for a 256MB Raspberry Pi but for any machine with RAM measured in GB it should be more than fine. 21:03:40 rwp: many VMs have RAM measured in MB not in GB, because €€€ 21:03:41 Oh, on the Sendmail side I am a complete heretic and run Postfix without any hesitation. 21:03:57 i just disable it, and on my pkgbase systems, I uninstall it. 21:04:21 i am considering to make a cloud-init module to do that for me… 21:04:59 rtyler, My interpretation of SMR is that if I were archive.org (no connection) then I would be thinking it would be worthwhile to make SMR work. Write once only ever. Read many times. And AFAICT reading doesn't have the penalty. The penalty is all in the writing. 21:05:00 CrtxReavr: re portsnap: is that thing still alive? 21:05:07 hahaha, the resilvering that I started this morning of one of these devices must have been stopped by the system bercause the disk is such a piece of garbage 21:06:27 yeah, /var/log/messages confirms that it looks like the drive was taking so long to respond that the kernel removed the device 21:11:32 rwp: wanna buy some 2TB drives? mildly used :P 21:12:06 meena, I looked at the memory use of ntpd on my amd64 system and found this https://bsd.to/RSA0/raw 21:12:07 Title: RSA0 21:12:53 rtyler, Yes on the kernel timeouts problem. That's a well known problem for SMR drives. And also for the old WDC Green drives too. Because they might take so long to respond. 21:13:56 rtyler, If you haven't read about TLER and ERC then please read about it here: https://raid.wiki.kernel.org/index.php/Timeout_Mismatch 21:13:57 Title: Timeout Mismatch - Linux Raid Wiki 21:14:57 I just looked at the receipt for the drives from January 2021, they were not listed as SMR drives. So that's fun 21:16:20 having realised how much all my disks and servers cost, I'm probably going to consolidate everything down to a single box with mirrored 20TiB disks 21:16:28 Curious... How many drives are we talking about? 21:16:40 I just want to know how much I should cry along with you about it. :-( 21:16:51 anybody have strong opinions to share on that? ( just 2 disks) 21:17:19 rwp: 4 drives 21:17:40 https://www.truenas.com/community/resources/hardware-recommendations-guide.12/ btw is my go-to 21:17:41 Title: Hardware Recommendations Guide | TrueNAS Community 21:18:28 Oh that's not so bad then. You get to keep paying your house mortgage okay then. And honestly those drives can still be used for write-once read-many archive storage needs. Just not in a conventional RAID. 21:18:31 I have some un-recommendations on some drives XD 21:18:38 dch, I have been making too much noise here but two spindles will be limited to 6 Gb/s transfer rates for 20T drives. 21:19:00 For what sounds like a personal archive with only you accessing the data I think that should be just fine. 21:19:07 rwp: this is the classic photos / music backup for a family so that is fine 21:19:31 But let's say you had an engineering lab with many people hitting the storage array then that 6 Gb/s becomes a very small straw. 21:19:45 and when the power crisis resolves itself I will sneak in another 2 and move the mirror to a striped setup like i currently have 2 deployments thereof 21:20:10 i really want fibre to the cellar and then I can use iscsi from my desktop 21:20:51 Seems reasonable. That's just my opinion and interpretation. 21:20:52 As drives became larger some people ran into performance problems. Making a backup for example can change from a short time to a multi-day backup run! 21:21:40 This year the city fiber project finally became available to my house and I admit to being completely spoiled by it now. 21:22:05 From my house to a friend's on the other side of the continental divide I am getting 890 Mbps data transfer rates. Wow. 21:22:12 rwp: https://bsd.to/NZXf/raw here's the openbsd daemon, not sure how that fares, cuz i'm pretty blind to numbers… 21:22:13 Title: NZXf 21:23:34 rwp: here's what I'm considering, curious to your thoughts: I can either buy 4 2TB drives, and replace them piecemeal in this array (RAIDZ2) or I can buy a couple drives (maybe 3x3TB), zfs send some snapshots into the cloud, re-image the machine, and then zfs recv them back 21:23:37 meena, I am confused by VSZ the virtual size being smaller than the RSS resident set size. 21:24:31 rwp: can confirm, I did a backup recently and it took over 24 hours just rsyncing the raid share to an archive disk (CMR even) 21:25:34 rwp: according to the man page, they are in kilobytes, btw 21:26:06 on both, FreeBSD and OpenBSD 21:26:27 Right. It makes sense though that a *HUGE* 4TB total array some time ago might have been 16x 512G drives. 16 interfaces. Now it would be only 2. 21:27:03 meena, Oh! I thought they were still in integer pages. My bad. Relatively though the numbers should still compare. 21:27:25 rwp: yeah, you're just off by 4, not any orders of magnitude ;) 21:27:25 But in that case ntpd on my FreeBSD system is only using 6MB of space not 24MB. Even better! 21:27:32 * meena still likes smol numbers tho 21:29:39 That OpenBSD ntpd seems to be running 3 processes. In which case you would need to add up the totals of all three. 2924+2772+1704=7400 which appears larger than the 5892 that I see. Hmm... 21:32:09 rwp: yeah, 7 instead of 6, again, we're in the same order of magnitude tho ;) 21:32:43 * meena has learned one thing about math as "engineer" and that is: you need to get the order of magnitude right, everything else, you let computers do 21:33:16 ** except when arguing with other engineers, then you have to get everything right 21:33:28 I am thinking that is listing unique memory pages too. I don't think it is double counting shared pages. 21:35:12 I wouldn't be worried about 5892 versus 7400 either but I was teasing meena a little bit because of the comments about small being important. But I think it is noise difference too. 21:35:59 * rwp must go afk to run errands... 21:36:01 rwp: when i say small numbers i mean: 5 or 7 vs 5892 or 7400 21:36:16 rwp: my head can't do numbers very well, and the longer they get the worst i get at them 21:36:33 Any idea how to fix this (trying to compile the kernel, FreeBSD 13.1, i386): https://pastebin.mozilla.org/LGDaFdDy 21:36:34 Title: Mozilla Community Pastebin/LGDaFdDy (JavaScript) 21:38:23 meandrain: on what platform? with what config? 21:38:30 i385 21:38:35 *i386 21:39:11 meandrain: which OS? version? etc… 21:39:20 this is the config file: https://pastebin.mozilla.org/UB7RyLAj 21:39:21 Title: Mozilla Community Pastebin/UB7RyLAj (JavaScript) 21:39:26 FreeBSD 13.1, i386 21:42:39 the config file is the one for AMD GEODE CPU, and it taken from nanobsd ( /usr/src/tools/tools/nanobsd/pcengines/ALIX_DSK ) and I've added "tuntap" and "wlan_acl" devices 21:43:24 meandrain: and that error then is telling you these devices need some other options enabled that nanobsd doesn't have 21:45:20 I've found some info here: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=241047 21:45:22 Title: 241047 – iflib: fail to compile kernel on FreeBSD 12.1-BETA2: Fatal error: ifdi_if.h file not found. 21:45:31 meena: thank you, make sense 21:45:48 meandrain: ifdi_if.h is part of iflib 21:46:05 yes, I'll add that and try, thank you again! 21:50:23 meandrain: so you need, according to iflib(4), device pci, device iflib, and i think that's it? 21:51:36 yes, it worked! >>> Kernel(s) BRIDGE built in 112 seconds, ncpu: 6, make -j6 21:52:13 meandrain: good luck booting it :P 21:52:36 ;) 21:52:51 * meena → bed 22:02:34 Before I get on my bike and ride I wanted to say that meena should be feeling very good about calling the problem of stealth SMR drives earlier. 22:02:39 That Seagate appears as a good example of a stealth SMR drive! 22:05:54 rtyler, A friend of mine just bought a box of 22 512GB SSDs for $200 off eBay. 22:06:11 The catch? They are all drive locked. That was known at time of sale. But if he can unlock them he will have 11T of SSD for US$200! 22:06:16 I tease him that he now has a gambling problem. But he thinks he can unlock them, just hasn't been successful yet. 22:38:32 why doesn't the sockstat command show which connection is incoming and which connection is outgoing? 22:43:39 if I type "sockstat -c", I can't see which connections originated locally and which connections originated remotely. 22:43:49 why not? 22:49:24 rwp: xD 22:53:31 which command should I type to find out which connection originated locally and which connection originated remotely? 22:57:00 Okay, I'll join some other irc channel to find out the answer to this question. 22:58:40 Oleg: once a socket is established that information isn't available 22:58:47 as i recall 23:18:31 it looks like this is related to my question: https://www.quora.com/How-can-you-tell-which-established-connection-is-incoming-or-outgoing-in-netstat 23:18:32 Title: How to tell which established connection is incoming or outgoing in netstat - Quora