Evilham | gnarface: that's user blaming, not cool | 00:03 |
---|---|---|
gnarface | that's not my intent whatsoever | 00:04 |
gnarface | i'm just trying to assess how critical the issue actually is | 00:04 |
redrick | I need a scorecard. Is the bone of content Debian-stable being 'in a non-existent state' (whatever that refers to)? | 00:06 |
fsmithred | no, the issue is that the OP was using "stable" in sources.list and it stopped working recently | 00:06 |
gnarface | i failed to parse that sentence, redrick, can you rephrase? | 00:06 |
fsmithred | still works with "ascii" | 00:07 |
redrick | Ah, sorry to late to the party. Carry on, then. | 00:07 |
Evilham | Well, it's critical in that behaviour changed in an unannounced, undocumented and unexpected fashion. It's not critical im that it has solutions and it doesn't breake *everyone* | 00:07 |
Evilham | So, "just use X" is harmful and u helpful | 00:07 |
gnarface | i disagree that providing a temporary stopgap of this sort is harmful or unhelpful | 00:08 |
fsmithred | well, in this case "just use X" is what it says on the website | 00:08 |
* redrick adopts 'u' as a shiny new prefix. | 00:08 | |
golinux | ^^^ Evilham nailed it. We have cowboys mucking around infra without communcating. | 00:08 |
gnarface | i do agree someone should address whatever is going on... | 00:08 |
fsmithred | yup | 00:08 |
golinux | They should let the team know when they are going to do major surgery. | 00:09 |
gnarface | but even Debian has long since years ago advised to use the release names in sources.list, because the very defined nature of the behavior of the "stable/testing/unstable" links is that they do change without warning | 00:09 |
redrick | Cwboys mucking around infra without communcating would appear to be the real topic. | 00:09 |
Evilham | gnarface: it is useful if accompanied by "there is an issue, sorry,use this instead" | 00:09 |
golinux | Bingo, redrick | 00:10 |
Evilham | Otherwise it's implicit user blaming | 00:10 |
gnarface | well i still disagree | 00:10 |
* redrick , like a stopped 24-hour clock, manages to be right once a day. | 00:10 | |
gnarface | i could have kept my mouth shut and that would have been less helpful | 00:10 |
gnarface | because there's still a chance someone didn't know you could use release names | 00:10 |
gnarface | or in fact that you SHOULD use them | 00:11 |
golinux | If user doesn't want to accept Devuan's default suggestions, failure falls on the user. | 00:11 |
redrick | FWIW, I think either release names or names of tracks are justifiable provided the admin understands implications, and can see gnarface's point about tactical utility of what he said. | 00:11 |
Evilham | Exactly | 00:11 |
gnarface | propagating the idea of using the stable link without including the knowledge that it's subject to change without warning is also harmful, IMO. i was just trying to prevent that. next time i'll just keep my mouth shut | 00:11 |
Evilham | Not what I meant, gnarface. You do great work, just when something is not working as it should, it has to be acknowledged. | 00:13 |
golinux | The problem isn't changing without warning as much as mixing two different releases in our case because we are always lagging behind debian. | 00:13 |
Evilham | And you can indeed try to teach better | 00:13 |
fsmithred | in addition to changing without notice there's also the possibility of getting debian versions of packages. (at least that did happen in jessie to ascii) | 00:13 |
Evilham | But not acknowledging issues is not useful, | 00:14 |
golinux | There are valid reasons to use the release name. | 00:14 |
Evilham | In anycase, I'm out of this topic indefinitely \o | 00:14 |
golinux | No one is not acknowledging that there is not an issue. | 00:14 |
redrick | A bad thing happening in Devuan because someone mucked with infra sounds distantly familiar, like it might have happened once before. Might be my imagination. | 00:15 |
golinux | It's just that none of us can fix it and those who can think they are above communicating with the community. | 00:15 |
golinux | They do what they want in their ivory tower of power. | 00:16 |
golinux | I've about had enough. | 00:16 |
redrick | Communication is a hard problem. Good place to start is be friendly & of good cheer, and assume good faith. | 00:16 |
fsmithred | I have to go to the kitchen | 00:17 |
fsmithred | bbiab | 00:17 |
golinux | This has been going on for weeks and communication isn't getting any better | 00:17 |
redrick | As with every other community technical project I've known, FWIW. | 00:18 |
golinux | When is the last time you saw parazyd post anything in this channel? Or on any of the channels? | 00:19 |
redrick | An important point, yes. To be serious for a moment. | 00:19 |
golinux | Maybe #devuan-arm - to which I don't subscribe. | 00:19 |
golinux | So much for "teamwork". | 00:20 |
golinux | Time for food here. ttyl | 00:21 |
redrick | As I'm married to an only child and her mother, also an only child, lives in my house, I'm familiar with the 'collaboration is for other people' concept. | 00:21 |
EHeM | You would need to live in the universe where Spock has a beard to be unaware that release names are *strongly* encouraged, yet used with care "stable" is still a handy tool. | 01:07 |
systemdlete | The deeper I go into understanding the raid subsystem and disks, the more questions I come up to. See https://raid.wiki.kernel.org/index.php/Drive_Data_Sheets#Non-Raid_drives_3 -- why is a certain make/model of disk drive considered raid or non-raid? My thinking is that ANY disk could be used as raid or non-raid. Please disabuse me of this notion. | 01:52 |
gnarface | it probably won't matter unless you're really measuring performance i think | 01:55 |
gnarface | my vague understanding of it is just that RAID will constrain the usage to very different and more narrow metrics | 01:56 |
gnarface | so some drive features matter more or less than others, compared to a single drive desktop | 01:57 |
gnarface | at least with software arid, for sure it won't matter what type of disks you use | 01:58 |
gnarface | software raid* | 01:58 |
gnarface | i don't really know if certain specific raid controllers might have specific disk requirements | 01:59 |
gnarface | it makes sense that they might | 01:59 |
gnarface | performance can vary by stripe size and disk performance | 02:00 |
gnarface | burst vs throughput performance | 02:01 |
gnarface | etc | 02:01 |
gnarface | at different write sizes | 02:02 |
golinux | EHeM: Except where there are two potentially conflicting "stable" releases existing at the same time. | 02:02 |
redrick | golinux: My favourite line from 'The West Wing': 'Ah, the rare valid point.' | 02:03 |
golinux | It has been reiterated here several times but EHeM knows better. :rolleyes: | 02:04 |
golinux | Time to take out compost. Just pulled up a bunch of rhizome propagating field grass. Oh joy! Worked up a good sweat and appetite. | 02:06 |
systemdlete | Looks like I stepped in at a bad time with my relatively unimportant issue. I think the issue of clear definition of "stable" (based on my reading of the comments above) is far more urgent. | 02:06 |
redrick | I have to keep remembering that most of my old-fogy knowledge is old-fogy Debian-specific knowledge and that there are very valid Devuan exceptions. | 02:07 |
golinux | We need a change of subject actually. | 02:07 |
systemdlete | sounds like people just need to air a grievance. | 02:07 |
systemdlete | It's relevant, esp. in this channel. | 02:07 |
redrick | Compost is endlessly discussable. ;-> | 02:07 |
golinux | Not in this channel | 02:07 |
systemdlete | in the debianfork maybe, not here. | 02:07 |
systemdlete | what golinux said | 02:08 |
* redrick was joking, I say joking. | 02:08 | |
golinux | +1 to systemdlete | 02:08 |
systemdlete | +1 to golinux | 02:08 |
* systemdlete tag teaming with golinux | 02:08 | |
redrick | +0 to redrick, to balance the scales | 02:08 |
systemdlete | no, -1 to balance | 02:09 |
golinux | Move to debian-fork in the wind? | 02:09 |
systemdlete | 0 won't do crap | 02:09 |
EHeM | systemdlete: At the top of that page 'Drives are classified as "raid" or "non-raid" depending on whether they support SCT/ERC.', more information on SCT/ERC can be found at https://raid.wiki.kernel.org/index.php/Choosing_your_hardware,_and_what_is_a_device%3F | 02:09 |
systemdlete | Ah, but that would be hardware raid, right? | 02:09 |
redrick | systemdlete: #ThatsTheJoke | 02:09 |
* systemdlete finally gets the joke | 02:09 | |
redrick | FWIW, I've used practically any old decent enterprise drive for MD RAID and not yet regretted that. | 02:11 |
fsmithred | A friend of mine was bored at work one day, so he and his co-workers made a raid array of usb sticks. | 02:11 |
systemdlete | I've used probably 8 drives or so, so far | 02:11 |
redrick | As that page says, e.g., WD 'Green' drives are examples of 'No, really, not enterprise.' There are others. | 02:12 |
systemdlete | I wonder if I need new glasses, or I have just become TOO adept at skimming. Thanks, EHeM. Shame on me for not taking the time to read ALL of the page, at least the intro material. | 02:14 |
systemdlete | Looks like at least one of my disks, a barracuda drive might be non-raid. But I am using software raid, how would a hardware feature impact this? | 02:15 |
* systemdlete reads on | 02:15 | |
* redrick is part of today's Lucky 10,000 to learn the specific term SCT/ERC, but previously avoided disaster by eschewing cruddy desktop drives for anything serious. | 02:15 | |
redrick | Seagate Barracuda in my experience tends to be non-crud. What model? | 02:16 |
systemdlete | oh. I see now. | 02:16 |
systemdlete | 7 second limit vs endless hang waiting on the drive. | 02:16 |
redrick | Yeppers. | 02:16 |
redrick | In WD's dictionary, 'green' means 'customer doesn't really care about ensuring data integrity'. | 02:17 |
systemdlete | ST3000DM001, but mine is a 1T, so maybe ST1000DM001, and I'm not even using that drive in my raid now. So maybe not an issue | 02:17 |
systemdlete | save planet, not data? | 02:17 |
redrick | Either of those is pitched as a desktop drive, but they look to a quick-eyeballing OK for RAID, AFAIK. | 02:19 |
fsmithred | please explain for mere mortals | 02:20 |
fsmithred | limit vs. hang | 02:20 |
redrick | fsmithred: If half a drive pair is waiting for the next ice age to actually write out data, that creates a problem upon power loss, etc. | 02:21 |
fsmithred | oh, got it. | 02:21 |
EHeM | The problem with installing a hypervisor on a new server is you start finding uses for any excess processor or memory; first my server got replaced simply due to running out of memory (had been on a machine with a mere 384MB, but the hardware still worked!); now it has a fileserver VM and an Android calendar/contacts backup VM. | 02:21 |
fsmithred | write now, sleep later, please | 02:21 |
systemdlete | Apparently, when the TLER feature is enabled, the drive is guaranteed to return in 7 seconds rather than hang possibly forever -- I think that is what is to be inferred. | 02:21 |
redrick | systemdlete: ja | 02:21 |
fsmithred | and what information in the smartctl output is relevant? | 02:24 |
plasma41 | I migrated my home directory from a single ext4 partition to a two disk mirrored btrfs volume this week. | 02:30 |
EHeM | The pages state `smartctl -l scterc` should give the information, but I suspect that requires a version of `smartctl` newer than what is in stable/ascii. | 02:31 |
plasma41 | Unlike with standard raid, I could write garbage data to one of the disk and btrfs would be able to detect and fix it. | 02:32 |
systemdlete | EHeM, fsmithred: On Ascii, I installed smartctl (gsmartcontrol) and I see sct listed at the end if I pass "-c" | 02:32 |
plasma41 | I usually run the gsmartcontrol frontend to smartctl instead, but when running smartctl directly, the '-a' option outputs everything IIRC. | 02:34 |
EHeM | Looks like those pages fail to document a crucial factor, I'm guessing SCT/ERC is an IDE/SATA feature, not something SCSI/SAS drives need support for. | 02:36 |
fsmithred | cool, it's there. | 02:40 |
fsmithred | WD Black | 02:41 |
EHeM | Isn't there, but I doubt it matters for a ST3000NM0023. | 02:44 |
systemdlete | https://raid.wiki.kernel.org/index.php/Choosing_your_hardware,_and_what_is_a_device%3F#TLER_and_SCT.2FERC "If it's available this feature needs to be enabled." OK, how? I don't see the mechanism in smartctl. Or is this somewhere else in the system? | 02:53 |
systemdlete | 2 of 4 of my drives support sct/erc. | 02:53 |
systemdlete | oh, I see. It's those ,N,M suffixes... | 02:56 |
systemdlete | so if I don't read EVERY LAST WORD OF EVERY LAST OPTION of some particular man page, I could overlook the solution. | 02:58 |
fsmithred | you can search in man pages | 02:59 |
EHeM | Tried looking at the page linked from there, https://raid.wiki.kernel.org/index.php/Timeout_Mismatch | 02:59 |
systemdlete | under the -l option, the man page says the following PRINTS either the smart error log, or the self-test error log, blah blah blah. It does not say in that intro to the -l option that some of the sub options allow you to set the values. So, I went to look at other options instead. | 02:59 |
systemdlete | Yes, EHeM. Got it already. LOL. It would be nice if the man page were a bit more... I don't know how to say it. Better laid out? idk | 03:00 |
systemdlete | Yes, the info IS there. But too much digging to find it. Better layout would be nice in many unix and linux man pages | 03:01 |
systemdlete | cat /sys/block/sd?/device/timeout: All 4 drives show 30 (3 secs) | 03:04 |
systemdlete | or is that really 30 secs... hmmm | 03:05 |
systemdlete | maybe not the same scaling | 03:05 |
systemdlete | the short script on that page suggests 180 as a value. So is that 3 minutes? 18 seconds? | 03:06 |
systemdlete | It's seconds, per an article about scsi drives on ibm's site. But that prob applies to ati also | 03:07 |
systemdlete | ata | 03:07 |
EHeM | Funny thing is for Linux ATA disks no longer exist, everything plugs into the SCSI subsystem which was much better written than the IDE subsystem was, so when SATA showed up they decided the SCSI subsystem was easier to interface with. | 03:26 |
redrick | EHeM: It'd be more accurate to say that libata author Jeff Garzik decided that leveraging the SCSI layer made it easier for him to code libata, which drives many PATA and SATA ATAs. And also SATA deliberately borrowed SCSI features, to the point that the next-generation SCSI, SAS, could then be created as just a compatible SATA superset with better-developed device addressing. | 04:49 |
* redrick confesses to being a SCSI bigot in dinosaur days. | 04:50 | |
* redrick kept a page about Linux support of SATA that for a decade was the primary reference on the Internet, which scared him since he didn't then use SATA, only collect data on it. | 04:51 | |
redrick | Anyway, libata was such a success and was so consistently well maintained where older ATA drivers were not that eventually all the HBA drivers that used /dev/hdX addressing became (to my knowledge) obsolete and (AFAIK) got dropped from the kernel. Which, et voila, brought about the eclipse of non-SCSI addressing of mass storage. | 05:27 |
redrick | For a SCSI bigot like yr. humble correspondent, that's a happy ending. | 05:28 |
Venker | hi people, good morning | 10:29 |
nemo | /bin/sh: 1: compress: not found | 16:41 |
nemo | anyone have any idea at all how to enable compress w/ debian tar? | 16:42 |
nemo | (that's for a tar -cZf | 16:42 |
nemo | ) | 16:42 |
KatolaZ | apt-get install ncompress | 16:43 |
nemo | ah thanks. was running through synaptic results but just getting confused | 16:55 |
furrywolf | am I the only one who doesn't like synaptic? | 17:06 |
furrywolf | I find it to make doing something things annoyingly complex. | 17:07 |
nemo | furrywolf: one thing synaptic breaks very very very easily is installing alternate versions of packages | 17:28 |
nemo | furrywolf: apt-get -t stable-backports is basically the only sane way to do it | 17:29 |
nemo | apart from that I rather like synaptic, and I'm a commandline fan | 17:29 |
nemo | but maybe I just am not familiar enough with debian tools for searching packages | 17:29 |
nemo | furrywolf: what would you use, for example, for a sortable search based on keyword of package name or description or contents that includes both installed and not installed, sortable, with status | 17:30 |
nemo | hm. I said sortable twice 😃 | 17:30 |
nemo | it's important to me 😉 | 17:30 |
nemo | being able to then select the ones for uninstall/install easily would be nice, but could just copy and paste I guess | 17:30 |
Hund | furrywolf: Is there something that you actually like? | 22:05 |
buZz | havent seen it | 22:06 |
golinux | Me either | 22:06 |
golinux | Wrong channel for this discussion though . . . | 22:06 |
buZz | hey golinux! how was your amsterdam trip? | 22:06 |
buZz | did you have fun? | 22:07 |
furrywolf | I like icewm. :P | 22:07 |
furrywolf | bbl, time for work. I like money too. | 22:07 |
golinux | pm | 22:07 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!