00:00:44 6I'm still running 4.13.17, pkg upgrade has not pulled in a newer one. 00:05:20 BraveheartBSD: i don't think it will 00:05:37 that's my problem. samba413 and samba416 are different pkgs 00:05:58 Message from samba413-4.13.17_8: 00:05:59 ===> NOTICE: 00:05:59 This port is deprecated; you may wish to reconsider installing it: 00:05:59 Reached its EoL on March 21, 2022. 00:06:42 It is scheduled to be removed on or after 2023-12-31. 00:07:05 so at the end of this year the pkg will no longer exist .. i'm guessing in favor of the samba416 pkg 00:07:41 sorry. thought this was the matrix # 00:10:41 Hmmm, I'll have to check it, don't think 4.16 is suitable with Win 11 00:15:10 ah. i'm using it for linux and macos (and i guess a win10 vm) 00:15:40 but my issue is that it's joined to an AD and it's a NAS but don't know if i'm supposed to install over 413 or remove it then install 416 and pray 00:16:25 i mean it seems like i SHOULD just be able to uninstall, install ... since i'm sure most of the background auth stuff is kerberos .. but i'm not sure. 00:16:48 i don't want to kill the NAS and spend forever sorting it out .. but i also want to upgrade before the pkg is removed 00:17:39 Yeah, mine is just a home fileshare 00:54:09 Macer, If you are worried something is going to "go away" you should backup your local copy from /var/cache/pkg/* so that you don't need to worry. 00:54:56 rwp I don’t think that is the issue here heh. 00:55:17 I just want to avoid it losing the AD then having to jump through hoops to rejoin. 00:55:44 It’s not really a pkg issue so much as it is a two packages of the same application issue. 00:55:56 I think that if you were to accidentally lose everything about ActiveDirectory that you would be better off. 00:56:32 🤔 01:16:33 <_xor> Macer: If you have ZFS, take a snapshot of the system and then upgrade and see what happens. If it fails, then rollback the snapshot. 02:03:56 _xor: ah ok. just didn't want some AD tomfoolery. i mean i'm sure if i have an issue i can most likely just remove the server from AD and re-add it 02:04:11 but i sure would like to avoid that. i'll save it for the weekend where i can have additional time to make up for any issues if they occur 02:05:10 <_xor> I haven't messed with AD in years, but I remember there being an issue with removing an AD from a forest and trying to add it back in (as in it wouldn't allow the server to re-assume the existing identity, it had to be added as a new server). 02:05:14 and who knows. maybe someone will chime in on the forum post 02:05:48 _xor: oh really? i've done it before using a samba dc (which is what i use now) 02:06:03 it was just a matter of removing it from AD then just re-adding it 02:07:02 in this case though it wouldn't be an issue if it were just a like pkg.. i'd just upgrade samba and be done with it.. samba is usually rahter good at not breaking things when you do major upgrades.. at least in my experience. but i guess freebsd separated it by version 02:07:12 <_xor> It'll probably be fine I'm sure, especially if it's a Samba AD. The issue I had was using Windows 2008. 02:07:19 <_xor> *Windows Server 2008 02:07:22 i guess 4.13 was supposed to be 'stable' and 4.16 was the new stuff 02:08:22 the odd thing is i don't know if pkg will remove 413 if you install 416 02:08:48 a piece of me is curious what happens if i just install it heh 02:08:59 maybe it's all symlinks or something 02:09:32 ah guess not 04:52:58 not sure if this is the best place to ask this but anyone have "wierd quirks" with sed regular expressoins behaving differently in neovim/vim vs zsh ? 04:53:20 i can not figure out if i am using different sed modules from editor to command line.. which would show the differene in how sed works 04:54:46 voy4g3r2: how are you using sed in vim? 04:56:45 :%s/,/,\r/g 04:57:05 i want to find commas and add newlines, which in bsd sed.. i have to do \r and NOT \n 04:57:21 but i try the same thing with sed -e 's/,/,\r/g' 04:57:27 it goes yeah.. too bad, i am not going to behave 04:57:39 i don't think it's really "sed", just internal regexes in vim and those can be pretty different from sed (as in command line utility) 04:58:16 okay, well documentation makes me believe it is.. but i trust your response 04:58:32 maybe i should not of tested the shell in a vim session and try and transition to a shell script :) 05:00:02 sed -e 's/,/,\n/g' seems to work for me 05:05:14 okay.. that is just odd, i must be mixing up things 05:05:17 I don't really know vim but in the old vi one would use Control-v aka ^v to enter the next character verbatim and then enter an actual carriage return into the expresssion. 05:05:18 that also works for me :) 05:05:44 In sed the escape sequences are definitely the way to do it. 05:07:14 anyway to get these 8 steps i do in vim to one command works for me 05:07:20 If you are wanting to replace the , with a newline then tr , "\n" is the way to go. 05:07:56 ohhh i like that option better than sed 05:10:02 it will require a little more verbose commands but much easier to read than sed 05:10:44 sed also does tr stuff too. All of the good letters were used so that is the "y" command. | sed "y/,/\n/" 05:11:01 baby steps 05:11:06 Which is convenient when there are other sed things to do too because they can all be done in one program invocation then. 05:11:11 i was so afraid of sed on hp-ux box, i learned perl :) 05:11:18 that is how large my adversion to sed is 05:11:32 But perl is not standard on HP-UX. You would have needed to compile Perl first! 05:11:52 yes, my boss liked perl so we had that 05:12:06 he was surprised i "found it" 05:12:10 Perl is a good general purpose text chainsaw though. Useful skill. 05:12:28 i have long lost the ability to write that stuff but i can read it and it was powerful 05:12:57 my context of was, is based on my use and not on people using it daily 05:13:33 i find this as my "fun time" as i am office 365 junkie most days.. today i got the "pleasure" of looking at SAS code 05:13:40 my brain still hurts from that ugly mess 05:14:01 and i do not even program for a living, i need to be able to understand it so i can transition out of SAS as it is hella expensive 05:14:15 for literal one statistical model execute a year 05:20:05 SAS, S, MATLAB, I just hate needing to learn an entire new language to do something singular or small. R, PSPP, Octave, those Free Software alternatives will at least have some long term lifetime. But I still don't want to learn Latin just to be able to compute an equation. 05:20:57 what needs to be done is singular in the model but the calculations and methods to execute said model, along with programming it is NOT 05:21:09 the requirements spec i have to disect is over 200 pages.. 05:21:55 If it needs it then it needs it. Glad you are dealing with it and not me. 05:22:00 thankfully i understand the domain of the model and not so much just the language, so can stumble along 05:22:06 haha 05:23:00 that is what the rest of the team said, they wanted to hire 5 people to review this, to only have to manage their work.. it is like, why.. just sit down and do it.. if we need the 5 people after this analyze.. go for it 05:23:48 but that is enough of my vent.. it is why i am now learning how to convert vim stuff to zsh stuff to reduce steps.. 3 days ago it took me about 80 steps per file, i am now down to 5 shell script lines and i write 2 commands 05:23:52 so happy about that 05:30:32 voy4g3r2, Shell programs are very powerful. When you get a moment you should read this classic example showing the power of them. It's worth the read. http://www.leancrew.com/all-this/2011/12/more-shell-less-egg/ 05:30:33 Title: More shell, less egg - All this 05:48:19 it seems pf table files support 2 syntaxes, comma after ip all on same line or no comma 1 ip per line. either better than the other? 06:39:56 Ummm freebsd-announce mailing list was meant to be only for announcements and now there's chain letter spam on it? 06:45:51 wildeboskat: if you're referring to the mail from glen, this may be useful context: https://lists.freebsd.org/archives/freebsd-stable/2023-November/001702.html 06:45:52 Title: Update to the Release Engineering Team Roster 06:46:30 Oh ok it sounded like chain letter spam, I don't know who glen is :) 06:46:42 I thought this mailing list was just meant as notifications for errata norices etc 06:47:53 Admittedly I have very low tolerance for anything that remotely sounds like spam. I've even gotten in trouble with work for lashing out at external sales people who emailed me out of the blue lol. 06:48:16 So it's probably me, if there's really something going on I'm sorry for him <3 06:50:04 Though I still don't think this is the place for that but anyway 06:50:41 I was only referred to said mailing list here last week and this is the first mail I got on it, that's why 06:50:49 my apologies 06:55:25 wildeboskat: not to worry, this is a very unusual occurrence 06:56:32 Thank you. I do know what it's like to be in a dark hole unfortunately 06:56:53 So I wish him all the best if he is here listening 06:57:14 dark hole is half-black or half-white hole? 07:04:48 half-black in this case for the people in it I guess :) 07:06:02 hello. can i mount openbsd crypto volume from my freebsd install? 07:12:48 is bioctl available on freebsd? 07:35:50 V-T60: unlikely, no 07:36:18 wrt bioctl, I find no mention of it in /usr/src 07:36:29 anybody familiar with ssh proxycommand and jumphost ? 07:39:56 I can ssh to A, and then ssh to B, but I cannot get ssh -J A B to work 07:40:32 oooh my bad, the end box is too old to support ed25519 keys 07:41:58 dch: hey! i lost the bookmark for DIY jails slides/presentation and cannot find on the web 07:43:18 angry_vincent: dobry rano! https://people.freebsd.org/~dch/talks/diy-jails/tutorial.html#1 | https://git.sr.ht/~dch/diy-jails-tutorial/ 07:43:19 Title: Jails from Scratch 2023 07:45:31 thnks much! 07:48:04 dch: we also say "dobry rano" here... 07:48:21 V-T60: then dobry rano for you too 07:48:39 dch: thanks 14:31:30 mind blown.. i setup NFS with /etc/exports and then find out zfs can do it WITHOUT making exports file 14:31:49 zfs likes to do its own thing 14:32:42 plus i do not have to keep reminding myself to do service mountd reload when i add a new export 14:32:45 works for me 14:35:57 zfs handles NFS ? 14:37:12 zfs can export its own datasets 14:37:37 see zfs share, etc. 14:43:32 It uses nfsd and /etc/zfs/exports, which is a special file that shouldn't ever be touched by the user. 14:46:06 mns: https://klarasystems.com/articles/nfs-shares-with-zfs/ 14:46:07 Title: NFS Shares with ZFS | Klara Inc 14:46:20 explains it pretty well. 14:46:47 and definitely highlgiths what debdrup said.. do not touch config files yourself, that is a no bueno moment 14:48:06 The only file you'll need to touch is /etc/exports and that's just to add "V4: /" on the first line :P 14:48:19 ohhh 14:48:27 that could be why i am getting permission denied 14:48:38 No, it just enables NFSv4. 14:49:18 zfs set sharenf'on, --alldirs, 192.168.0.115' storage 14:49:26 zfs set sharenfs='on, --alldirs, 192.168.0.115' storage 14:49:38 hrm. we want nfsv4, so adding that right now 14:49:50 voy4g3r2: thanks 14:50:28 Please see the note in zfsprops(7) about comma separarated values. 14:50:59 hehe.. stumbling along this journey.. just with random zeel 14:52:59 Some asshole had to go add a note about that: https://github.com/openzfs/zfs/commit/28401f6668 14:53:00 Title: zfsprops.7: Add note about comma-separation · openzfs/zfs@28401f6 · GitHub 14:55:42 hrm.. yeah is aw that in the blog entry.. make sure you use commas, i guess i have to make it look ugly.. by smashing it all together 15:25:19 out of left field, for any of you using Roundcube, it was just announced that they've merged with Nextcloud 15:40:30 mns, did you end up having the same issue with the new version of bastille? 15:42:48 scoobybejesus: I did not. I am on an older version of bastille. Also I'm looking at appjail as well. Only because I want to try out different things :-) cbsd seemed a little more complex compared to bastille. 15:44:06 appjail seemed interesting when i first saw it, but i have not dug in. yeah, cbsd seems maybe like what i'd want for easy vm's, though I think i'd rather learn vm-bhyve better first 15:45:47 scoobybejesus: yeah if I needed orchesrtration for a production environment, I'd use cbsd. This is for home stuff, so bastille suits the needs. appjail seems similar to bastille as well 15:49:32 mns: cbsd? 15:51:12 Demosthenex: https://cbsd.io/ 15:52:44 Demosthenex: here we go: https://github.com/cbsd/cbsd 15:52:45 Title: GitHub - cbsd/cbsd: Yet one more wrapper around jail, bhyve and XEN on FreeBSD platform. 15:52:55 says no zfs? 15:54:37 I don't recall. Its been a while since I touched cbsd. Also I'm at work and since its a .ru domain, its blocked unfortunately 15:55:27 yeah, i think it says its not zfs only 15:55:34 but has "generic" integration 15:55:41 interesting to read about 15:59:40 some discussion here on maybe Monday made me go 'I need to install 13.2 in a VM' but I forgot what for 16:00:15 you got logs? 16:00:24 probably 16:01:01 found it 16:26:39 hey 16:27:34 I'm trying to use libelf to read an ELF file. The code works perfectly under linux but when I want to use it under Freebsd, elf_begin(fd, ELF_C_READ, NULL) returns NULL and elf_errmsg returns "Invalid argument". 16:28:14 I don't understand since when I truss the tool libelf doesn't try to read the ELF file.. there's no read operationg.. 16:28:27 of course open returns a valid fd 16:28:53 and when I read the fd it returns the first bytes of the ELF file.. so it's ok 16:29:08 there's something I don't understand with the FreeBSD version of libelf :| 16:30:14 titou: what does elf_errmsg(-1) return? 16:30:20 because that code is valid 16:30:33 I don't know how can I debug that.. there's 3 lines...: open, elf_version(EV_CURRENT) and elf_begin 16:31:04 "Invalid argument." 16:32:28 titou: see the man page for ELF_E_ARGUMENT error cases and try to figure out which one it is would be my next step 16:32:48 i already check that page.. 16:33:09 but as far as I said there's just 3 poor lines of code 16:33:50 and the only thing which could help me in the elf_begin man is: "The value in argument cmd was unrecognized." 16:33:55 titou: what's the truss output? 16:33:57 right, but there's also a file you're trying to read. are you sure it doesn't contain some unexpected things? elf_begin(3) has 7 error cases for ELF_E_ARGUMENT that can happen 16:34:02 because other things are for a third element different from NULL 16:35:06 RhodiumToad: my open which returns a valid fd, and just after an error because elf_begin returns NULL. No syscall is performed by elf_begin 16:35:20 not even fstat? 16:35:26 not even! 16:35:39 what fbsd version? 16:36:09 dstolfa: i also forgot error when fd is for an empty file (but as I said, i can read this file...) and for unsupported file type (but it's a regular file...) 16:36:16 and you're sure your third arg to elf_begin is really NULL? 16:36:18 RhodiumToad: 13.2 16:36:28 titou: what do you mean by "regular file"? 16:36:33 is it a valid ELF file that other tools can manipulate? 16:36:47 titou: are you sure fd points to an elf file? 16:37:41 https://pastebin.com/0RcjAkkz 16:37:42 Title: int main(int argc, const char **argv){ Elf_Scn *scn = NULL; GElf_Shd - Pastebin.com 16:37:43 here is the code 16:38:01 dstolfa: nimaje: readelf and objdump can read and dump if yes 16:38:09 it* 16:38:18 what about FreeBSD's elfdump? 16:38:27 titou: and the truss output? 16:38:52 and you're sure you haven't managed to redefine NULL anywhere? 16:39:06 and how are you compiling and linking it? 16:39:14 dstolfa: it works very well.. the corresponding file is generated by gcc.. 16:40:19 RhodiumToad: https://pastebin.com/UdR3f3H8 16:40:20 Title: openat(AT_FDCWD,"debug/apos",O_RDONLY,00) = 3 (0x3)Unable to read the f - Pastebin.com 16:41:10 how are you compiling and linking the program containing the elf_begin? (not the one you're trying to read) 16:41:38 and what architecture? 16:42:19 g++ -o exportSym tools/exportSym.cpp tools/hashList.cpp -lelf 16:42:25 x86-64 16:42:42 hmm 16:42:58 have you tried with clang? 16:44:29 I'm wondering if this is a problem with enum sizes 16:45:00 indeed it works with clang.... that's so strange 16:45:07 i don't understand 16:45:09 I do 16:45:46 what's almost certainly happening is that something's messed up the calling convention, so elf_begin is seeing its third arg as not being NULL 16:46:05 that's so incredible 16:46:23 ah yeah, C++ doesn't have a well-defined ABI so you're hitting that 16:46:41 can you compile the relevant source file to .s using g++ -S and pastebin that? 16:47:05 the libelf funcs should be declared as extern "C" so c++ abi issues should not be in play 16:47:30 (libelf.h does contain the expected declarations for that) 16:49:18 you are including libelf.h properly, I hope? 16:54:10 https://hastebin.com/share/ivemehuros.perl 16:54:11 Title: Hastebin 16:55:15 that's with g++? is it any different with clang? 16:58:06 https://hastebin.com/share/kayuxenale.perl with clang 16:58:07 Title: Hastebin 16:59:31 the second argument is different... 16:59:34 5 vs 1! 17:00:50 that's so strange.... it seems to be a bug?! 17:01:25 and the register for the last argument is diffrent %edi vs %edx 17:02:53 the earlier pastebin (https://pastebin.com/0RcjAkkz) doesn't show any includes. i'm assuming there *are* in fact includes in your program? 17:04:07 and that the way you include libelf is via #include , which would result in all the function declarations, including elf_begin, to be under an extern "C" as RhodiumToad mentioned 17:05:16 5 is correct 17:05:38 so I was actually wrong 17:06:15 somehow you got ELF_C_READ defined as 1 rather than 5, so elf_begin is complaining that the second arg is wrong 17:07:30 nimaje: the registers are set up in a different order but that's not relevant 17:08:04 the args go in %edi, %esi, %rdx in that order 17:09:08 I'm going to guess that with gcc, you're maybe picking up a libelf.h from gcc/binutils rather than the freebsd one? 17:10:21 RhodiumToad: yes, that's it. i've just checked it. ELF_C_READ in elfutils' libelf is defined as 1 17:11:39 https://github.com/cuviper/elfutils/blob/08ed26703d658b7ae57ab60b865d05c1cde777e3/libelf/libelf.h#L140 17:11:40 Title: elfutils/libelf/libelf.h at 08ed26703d658b7ae57ab60b865d05c1cde777e3 · cuviper/elfutils · GitHub 17:13:46 ok, so it's compiling against a libelf.h for a different library than the one it ends up linking to? 17:14:11 so mismatch between -I paths and -L paths, basically 17:21:11 dstolfa: I confirm 17:22:03 hmm there's two version of libelf on FreeBSD ? 17:22:08 no? 17:22:23 no, it's because g++ uses its own instead of using the system's libelf defines 17:22:30 it links to freebsd libelf, but it uses the headers from g++ 17:22:47 I didn't find a libelf.h in the gcc package 17:23:04 you're right: i've two libelf.h: /usr/include and /usr/local/include and enum are different.. 17:23:12 shit.. 17:23:24 pkg which /usr/local/include/libelf.h 17:23:33 probably elfutils, tbh :D 17:23:37 it's elfutils 17:23:40 right 17:24:10 so if you compile against elfutils then you need to be sure to link against it too, not to the base libelf 17:24:13 ok so I'll try to play with -I... in order to get it working 17:24:35 or using the right library yes 17:24:41 note that base system clang does not look in /usr/local unless you tell it to 17:24:53 that's interesting 17:25:07 (whereas port gcc probably does, i haven't checked lately) 17:25:49 anyway, all is explained 17:26:06 is there a solution to force gcc using /usr/include rather than /usr/local/include ? 17:26:18 -I/usr/include doesn't do the job 17:26:50 there are a dozen variants on -I that add to different places in the list of include dirs 17:28:08 titou: you could pass -nostdinc and manually specify them 17:28:38 that's probably not wise 17:29:02 probably better to look at the actual list of dirs being used and figure out how to set it correctly 17:29:44 well yes if you're making a program you want to use more than just for a quick test, you should definitely fix it the right way 17:30:13 ah 17:30:24 i'm agree with dstolfa .. i want to get it working for different computers, not only mine 17:30:37 checking gcc docs, if a dir is already specified as a "system" include dir, then another -I for it is ignored 17:31:01 yeah but i don't know if -nostdinc is a good idea then. if this is supposed to build on a wide variety of platforms and setups, i would set up a proper cmake build around it 17:31:13 so you might in fact need -nostdinc 17:31:58 hm, now i actually wonder if cmake handles the libelf issue properly 17:34:21 i wonder if find_path(... PATHS /usr/include ... /usr/local/include) would handle this properly 17:34:34 is it possible to set "kern.elf64.allow_wx" setting for a single executable? 17:36:48 ngortheone: I think so, via a flag in the elf brand 17:37:41 ngortheone: yup, see the elfctl command 17:46:59 thanks RhodiumToad! 17:51:58 thank you so much for your help!! 18:27:47 <_xor> Can I use `poudriere jail -j myjail -u -m src=/usr/src` to install a source tree into an existing builder jail? I'd try it, but that system is currently building and I don't want to mess with it. 18:28:38 <_xor> I want to say that I remember trying that at one point and it didn't work. Only alternatives I could think of were to extract it into the dataset manually and re-create the @clean snapshot, or just re-create the jail. 19:35:44 is there a way to make FreeBSD idle better? (beyond powerd)? on an 11th Gen Intel i5-1135G7, after a clean FreeBSD 14 install, idle CPU temps are ~47C. Same system with a clean Arch (linux 6.6), idle CPU temps are ~33C. 19:37:06 markmcb, Yes. Install powerdxx from ports and run it instead of powerd. (Something I recently learned about too.) 19:38:04 rwp: thanks, i'll check it out 19:38:08 It turns out that powerd is an old program from FreeBSD 5 days that made it into base but has not had the love or attention needed to keep it working with newer systems. 19:38:46 So the powerd++ author really did a good job writing a similar but more powerful replacement that works with currently available systems. 19:39:04 _xor: you should not do that with a poudriere jail 19:39:15 just create a new one and copy the options directory over 19:40:10 markmcb, Hint: After installing powerdxx run it in the foreground and watch the verbose information dump. Then realize that you also need to install either "coretemp" or "amdtemp" depending upon your cpu model. 19:44:25 markmcb: what does sysctl dev.cpufreq.0.freq_driver give? 19:44:34 And by foreground I mean with both -v verbose and -f foreground. Here is an example: https://bsd.to/jNDA/raw 19:44:35 Title: jNDA 19:45:32 angry_vincent: est0 19:46:45 this is maybe strange. since gen8 there is new hwpstate_intel driver 19:47:29 it should be set for your cpu, but maybe it's not for some reason 19:48:01 angry_vincent: user error, lol, i looked on the system i'm on now without thinking. i'll have to check the other system later. 19:49:43 ok. then if it is hwpstate_intel0 for you, then you may drop powerd or powerdxx completely and use newer driver, as both powerd/powerdxx would not be correct. 19:50:06 markmcb: https://docs.freebsd.org/en/books/handbook/config/#hwpstate_intel 19:50:07 Title: Chapter 14. Configuration, Services, Logging and Power Management | FreeBSD Documentation Portal 19:50:26 markmcb: https://man.freebsd.org/cgi/man.cgi?query=hwpstate_intel&sektion=4&format=html 19:50:27 Title: hwpstate_intel(4) 19:56:19 correction: since gen 6, not gen 8 19:58:04 thanks! will read through this 20:01:15 https://social.treehouse.systems/@marcan/111493984306764821 20:01:16 Title: Hector Martin: "WARNING for Raspberry Pi 5 NAS users: **If you us…" - Treehouse Mastodon 20:01:38 whoops, wrong channel 20:02:10 that seems like a bad idea outright 21:17:54 angry_vincent: the system that's reporting est0 had the feature disabled in the BIOS. i enabled it and now see hwpstate_intel0. turned off the powerd options. will monitor and see. curious that it was disabled. all other cpu features were on by default. i wonder if there are downsides to using it? regardless, thanks! 21:25:36 looks like freebsd-update upgrading to 14.0-RELEASE from 13.2-RELEASE on digitalocean is predictable and easy enough. fingers crossed but everything seems to be in order 21:29:09 scoobybejesus: I hope you will not get teh vnlru weirdness Ihave 21:29:14 scoobybejesus: I've heard some pretty awful things about DigitalOcean and FBSD 13+. 21:29:14 I have* 21:29:22 though I am not on DO 21:35:52 I'm running a few 13.x on DO 21:37:25 This has been alive since 12.1. Getting to 13.0-RELEASE was rough because of having to install cloud-init on my own, since it wasn't part of DO's base, i guess. Otherwise, it has cooperated i think 21:40:13 I know someone that had some pretty strange file system issues after upgrading to 13.1-R. He assumed it was due to DO dropping support for FBSD. 21:42:48 I was under the impression that they just scrapped the existing FreeBSD install images and stopped making new ones 21:42:51 i am currently experiencing mount_nfs: nmount: /storage/blinksystem: Stale NFS file handle and the odd things is, the machine that is having it.. does not have a mount of this location. When I do some searching it is assuming i am using exports, which i am not.. i am using openzfs sharenfs option. Does anyone have a tip/trick to evaluate how a stale nfs file can occur on a client? 21:43:28 vkarlsen: I was under the same impression. 21:43:51 mount -t nfs 192.168.0.115:/storage/blinksystem /storage/blinksystem <-- this is the command that gives me the stale NFS file handler 21:44:26 Nothing seemed to changed for me. I also tested uploading my own homebrewed 13.2 image. 21:46:54 https://forums.freebsd.org/threads/random-bad-file-descriptor-errors-building-ports-on-a-vps-from-digitalocean.86485/ 21:46:55 Title: random Bad file descriptor errors building ports on a vps from Digitalocean | The FreeBSD Forums 21:51:19 that has happened to me many times. i attribute it to being some virtualization issue and inability to share resources smartly. i reboot and re-run the operation and it completes fine 21:52:23 my prod is basically just for me, so i can afford to reboot whenever i want. i can see how that would drive someone crazy, particularly if it only happens on DO 21:53:11 Hm, I have never seen that problem on any of mine 21:53:35 Maybe I'm in a lucky neighborhood 21:54:08 voy4g3r2: is something removing that directory and re-creating it? i know that causes stale handles 21:54:30 it tends to only happen when others "sharing" the CPU are hard at work (based on huge "system" CPU resources being used for basically no reason from my perspective) 21:54:40 i used to get in the habit of: cd `pwd`; on one set of servers to prevent it 21:56:36 my current VPS is a guaranteed resources thing and it is noticeably better than the VPS I tried before 21:56:59 I guess the model also matters, KVM now vs openvz before 21:57:51 discounting I/O the new VPS is just a bit more performant than my old colocated E3-1240 21:58:07 frequency is lower, but core count is higher and there's more RAM, yay 21:58:39 plus storage is 1 TB SSD rather than 2x2 TB spinning rust RAID1 21:59:01 but the best part is price 22:02:31 rtprio: i never did have it mounted, but tried that and still no bueno 22:02:56 <_xor> voy4g3r2: sudo nfsstat -m 22:03:11 throw back nothing 22:03:45 voy4g3r2: you mean you can't get it mounted with that mount command? 22:03:57 yes sorry, was not clear yes 22:04:47 can you mount it to a different directory? is something already mounted on /storage/blinksystem ? 22:05:39 <_xor> voy4g3r2: Are you using NFSv3 or NFSv4? 22:06:13 <_xor> Also, you can try setting sharenfs to none and then back to the current value, then trying to mount again on the client machine. 22:06:47 <_xor> I don't remember which daemon it is, but restarting one of them (depending on NFSv3 or NFSv4) should also flush that stale handle. 22:10:07 Type "nfs" is v3 and v4 is type "nfs4" AFAIK. The problems I have seen in cloud hosting with nfs are network related. NFSv3 uses UDP and if networking is not solid then NFS is not solid. 22:10:44 But on LANs I am heavily using NFSv3 and have had zero problems through FBSD 12 and 13. 22:11:32 uh,, using UDP and rely on stable networking sounds sketchy x) 22:12:17 the okay let me give this a try 22:12:28 It's a protocol designed for UDP from the start though. So if there is some packet loss due to $REASONS then things still work pretty well. 22:12:46 But if packet loss becomes extreme then of course things go sideways. 22:13:56 okey, then i follow. 22:14:19 Also there are stale handles and stale file handles and they are somewhat different things. 22:14:22 The main reason I see people running into stale handles is that the server has dropped out. 22:14:28 But stale file handles are almost always due to use in a way that tickles the file attribute cache across distributed systems. 22:14:49 my /etc/exports is blank, so i am assuming NFSv3 unless something else defaults it.. server is freebsd 140 22:15:13 ld-elf.so.1: /usr/local/lib/php/20200930/mbstring.so: Undefined symbol "php_pcre2_compile" wut 22:15:18 The /etc/exports is the non-zfs way to export things on servers. Servers with zfs use zfs to manage the exports. 22:15:29 And on clients it is in /etc/fstab 22:16:59 yeah, first testing it, to see if mounts than will do that 22:17:13 rwp: i am NOT going through that whole NFS issue i had last time, where since synology died i couldn't boot a machine 22:17:19 jbo: This looks similar to what I ran into last week: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=275361 22:17:20 let me try "refresing" this sharenfs stuff 22:17:21 Title: 275361 – devel/pecl-APCu: Undefined symbol "php_pcre2_match_data_create_from_pattern" 22:20:01 voy4g3r2, In that case maybe use unison or syncthing and just keep all of the data copied everywhere? :-) 22:20:23 showmount -e is showing what's expected? 22:20:45 rwp, wait, you can export NFS via ZFS (or the other way around)? I am currently setting up a backup server to which several hosts should push their backups to. I was thinking of using zfs send|recv over SSH. Is the NFS+ZFS approach more interesting? 22:21:24 no, you set your zfs to 'export' on nfs; akin to adding it in /etc/exports 22:21:28 mariuss, thanks for linking - I'm investigation. this is php80 tho 22:21:31 jbo, Uhm... I think you are asking several different questions all at once! 22:21:33 now i am confused. i setup sharenfs option on some pools, and i can us mount -t nfs to mount 22:21:38 still getting this damn stale thingie 22:22:25 rwp: i went through enough headaches just getting the synology -> openzfs, would like to keep it centralized to this one machine with all the mounts 22:22:29 rwp, rtprio, ah, so the client still sees/speaks NFS. it's just a "different way" of handling it on the NFS server? 22:22:34 Double check that there are not old mounts *underneath* the mount point of the new mount you are trying. Just a guess at brainstorming a problem here. 22:22:35 but to make myself NOT go insane, i am going to restart the file server 22:22:37 and strt again 22:23:39 jbo: yeah, pretty much 22:23:44 jbo, On the server side before zfs the file systems were exported using /etc/exports file. Since ZFS is somewhat of a total manager tool the contents of the /etc/exports were moved into a "sharenfs" attribute on a dataset so that one could use zfs for all of the management. I recommend doing it that way. 22:23:48 rwp: nope both client and server are both new installs 22:24:09 But the old way of using /etc/exports is still allowed. Okay to continue doing that if one wants to do that. But I am old school and I converted to the new way and recommend it. 22:24:13 https://klarasystems.com/articles/nfs-shares-with-zfs/ 22:24:14 Title: NFS Shares with ZFS | Klara Inc 22:24:22 jbo that may help a little to explain what rwp is mentioning 22:24:32 it shows the transition from /etc/exports to the sharenfs concept in zfs 22:24:41 jbo: i was going to use restic to backup but haven't gotten around to it yet 22:24:45 Meanwhile backups... If you are wanting to set up send|recv then using ssh is probably one of the simplest ways to do it. I would continue. 22:25:13 I don't in my head have a way to tie send|recv together with nfs mounting. Unless one is doing only send to a file but... Let's not do that. 22:25:43 alright, understood. 22:25:58 yeah I am still debating whether I want to do backups by just giving each host an NFS share or by just using zfs send|recv 22:26:08 there is the stale fix 22:26:11 it was the server :) 22:28:24 any thoughts/recommendations on the backup via NFS vs zfs send|recv situation? 22:28:34 rtprio, I am using restic and +1 it is a very good backup tool. 22:29:47 jbo, My thoughts are that for backup I would either use a backup structure like restic or I would use ssh+send|recv+zfsautosnaps 22:30:56 I myself don't have a backup process running with ssh+send|recv+zfsautosnaps, yet, but I am in the journey of setting one up, at a slow background pace. On my largest file server with a lot of data that changes in big lumped chunks. 22:32:05 It takes a couple of hours to do an rsync pass across that NAS. So I am going to be converting it to an ssh+send|recv+zfsautosnap backup. I haven't found anything premade that I like. So I am tinkering one for me together. 22:32:21 rwp, I looked at restic several times. it looked quite decent. not sure how to decide whether to go restic or the zfs send|recv route. I'm looking for a decent mechanism to backup servers to a local backup server 22:32:39 Meanwhile, on other systems, I am using restic to push encrypted backup blobs to remote backup servers and it is working flawlessly for me there. 22:33:41 And I am still using BackupPC for local backup in a couple of places. It's a central pull based system. Needs a trusted central system. Uses rsync. 22:33:50 I am already using zfs send|recv to push backups to a remote server. the main problem I am encountering is that I can't easily send only some datasets selectively. 22:34:03 Previously I have used a bunch of both typical programs and home made backup systems too. 22:34:57 jbo, "can't easily send only some datasets selectively": Could you expand on that some? I don't understand. 22:36:18 rwp, lets assume I have datasets storage/a, storage/b, storage/c and so on. so far I am create recursive snapshots on storage and then send that via zfs-send. however, I ran into the situation where I want to send storage/a and storage/c, but not storage/b. 22:36:37 in my scenario it's because storage/b is my poudriere repo 22:36:57 so traditionally I'd want a switch like --exclude-paths=storage/b but that doesn't seem to be an option. 22:38:13 rwp: i am having a brain fart, what was that option i should set on the client side (NFS) so it will boot if it does not see the mount? 22:38:27 First, I don't know. I haven't needed to do that yet so haven't fought that issue yet. But I think... Don't use the recursive option, list them all out in a script and exclude the one you want to exclude in the script and then send the remaining ones. 22:38:52 voy4g3r2, "bgnow"?? 22:40:02 <_xor> Gah 22:40:16 voy4g3r2, Yes. "bgnow" is what I am using instead of "bg" which avoids a 60 second delay. See "man mount_nfs". 22:40:30 <_xor> I guess I can't create a poudriere jail using pkgbase to actually create the jail AND also putting /usr/src in it. 22:41:17 _xor: mount nullfs? 22:41:34 ah yes, okay.. you mentioned that throught he lesson a few days ago and did not have in my notes, just had.. you should STOP client from NOT halting with /etc/fstab entries but did not get it down :) 22:41:49 <_xor> You mean via hook? I mean yeah that's an option, though I'd prefer not to have to do that. 22:42:32 <_xor> I just tried adding -S /usr/src to my poudriere jail command, in addition to -m pkgbase=file:///mnt/repos/freebsd/dists/1400501/pkg, and it didn't put the source in there. 22:42:33 voy4g3r2, I think that must have been someone else who suggested that in the last week because I don't remember it coming up from me in the last month. 22:43:10 oh, my bad.. someone did 22:43:25 my notes are bad, i should move these to onenote as my hand writing is horrible 22:44:47 It's fine. If we both suggested the same thing then it is even better! Confirmed in that case. :-) 22:45:17 exactly 22:45:21 The other option is to calm down when writing notes and to make them more legible? Just saying! :-) 22:45:42 haha, that is what the wife says for years... i also type way more than i write 22:46:03 i will take under advicement but probably still just be horrible.. as it is just bad 22:47:49 <_xor> All right, let's see if this works. I just destroyed the jail@clean snapshot, extracted the source tree into it, and created a new jail@clean snapshot. 22:48:38 <_xor> Oooh, nice it looks like it actually worked. 22:57:04 i think i had a hacked up synology nfs file server.. now i am getting grouping issues between both.. my user is in a group but it is not "passing" to the client 22:57:07 fun times 23:01:15 There is about a half dozen ways of handling groups over nfs, or so it seems. The original way was a fixed array of gids. But it is fixed in size. That is the most reliable though. So the new way is having the server also know about all of the group accounts. Which I don't like because it entangles the server in with the user accounts. 23:01:55 <_xor> sec=krb5 23:05:31 yeah well the server has the groups, the client does not 23:05:40 so it is like do i pull a --manage-id tingie or not 23:06:04 or do i create the same group on the client aht is on the server and do that way... my luck the gid won't line up anyway 23:06:49 <_xor> Are you the guy running Samba AD? I can't remember if that's you or someone else. 23:07:03 nope, not me 23:07:17 <_xor> If it's you, then you can enable krb on NFSv4 and use that with your AD to have a single mapping of UIDs/GIDs. 23:07:17 i was playing with afp and time machine backups and all that fun stuff 23:07:23 <_xor> oh 23:07:56 because my synology died 23:09:06 _xor, You are thinking of Macer who was talking about AD. 23:11:20 but must continue this later, dinner time.. thanks for the help as always, further along than i was an hour ago