| === Napsterbater is now known as Guest96179 | ||
| === Napsterbater_ is now known as Napsterbater | ||
| === Napsterbater is now known as Guest11261 | ||
| === Napsterbater_ is now known as Napsterbater | ||
| ruben23 | hi guys anyone can help how to add user or user email on my postfix mail server, anyone have idea.? | 04:32 |
|---|---|---|
| lordievader | Good morning | 08:30 |
| === denningsrogue6 is now known as denningsrogue | ||
| vlm | i made a raid 1 from two raid 0 arrays with mdadm,how do i go about scrubbing the raid,is it enough to scrub the main raid 1? | 15:56 |
| sdeziel | vlm: did you kind of re-implemented raid10 by stacking multiple md devices? | 16:00 |
| vlm | sdeziel: dunno maybe,ive never made a raid before though,thought this was how i could make a raid 10 or could i do it right away? | 16:28 |
| sdeziel | vlm: I got confused what you built is called a raid0+1. It seems that mdadm requires stacking like you did for this raid level. raid10 is similar (but inverted) and has native support by mdadm, no stacking required | 16:29 |
| vlm | sdeziel ah ok ,ill head over to kernel.org try get read the guide there and read up on lvls aswell thanks for asnwer | 16:30 |
| sdeziel | vlm: you should probably take a look at ZFS too. I've switched away from mdadm to ZFS and wouldn't look back | 16:31 |
| sdeziel | there are some special edge cases where mdadm is a better solution but for pretty much everything else, I feel ZFS nails it | 16:32 |
| vlm | sdeziel yeah i always wanted to give it ago,just i keep reading here and there that is not as stable or so as it is in freebsd and such? | 16:32 |
| sdeziel | vlm: I seriously question those claims. ZFS is in 'main' on Ubuntu, so you get good support for it. Been working well since 16.04 IIRC | 16:33 |
| vlm | sdeziel also i think with raidz i wanted i couldnt easily expand it without requiring much more hardware or so i think it got expensive | 16:33 |
| vlm | sdeziel oh thats nice to hear though,yeah i always wanted a raidz or so | 16:34 |
| sdeziel | true, growing a zpool requires some thinking. That's why usually go with simple mirroring that are all tied up into a stripe at the zpool level | 16:34 |
| sdeziel | this gets you a kind of raid10 on steroid that is easy to grow | 16:35 |
| vlm | sdeziel ill look into that,though this is just a simple raid i dont need snapshots for it so maybe ill be ok with mdadm | 16:36 |
| sdeziel | maybe you don't need snapshot but you surely want checksums to protect your data ;) | 16:37 |
| sdeziel | and transparent compression, and ... | 16:37 |
| vlm | sdeziel: though doesnt the mdadm scrubs do that ? | 16:37 |
| sdeziel | not really | 16:37 |
| sdeziel | or I should say, really not ;) | 16:38 |
| sdeziel | I don't have to go into details ATM but if you care about your data ZFS has more to offer natively than mdadm can do | 16:38 |
| vlm | sdeziel alright think ill consider trying it then if i can achieve the same setup 0+1 | 16:40 |
| vlm | sdeziel: i now made one vdev i believ consisting of 2 mirrors,that should be the same as 0+1 shouldnt it? | 17:38 |
| sdeziel | vlm: that's more akin to raid10 | 17:38 |
| vlm | sdeziel: arent thouse in effect the same or so? | 17:41 |
| vlm | sdeziel: they seem so similar to me -_- | 17:41 |
| sdeziel | vlm: I'm not an expert by any mean, I'm basically looking at https://en.wikipedia.org/wiki/Nested_RAID_levels | 17:42 |
| vlm | sdeziel: im seeing a noticable improvement in throughput now though,up 200MB/s + up from 184MB/s something | 17:42 |
| sdeziel | vlm: but ZFS is a different beast | 17:42 |
| sdeziel | vlm: while you benchmark your new array, check 'zpool iostat -v 1' | 17:44 |
| vlm | sdeziel: my initial reading tells me they are the same in performance and redundancy, just different layout also raid 10 might not be supported on ancient hardware maybe | 17:47 |
| vlm | sdeziel: my system is very low on ram so that might get challenging i tuned arc_max and min for starters now i need to read up on how to track down if there a memory issues | 17:48 |
| sdeziel | vlm: you can reduce the ARC usage if you really cannot afford the RAM cost... it will harm the performance of course | 17:50 |
| sdeziel | you can tune the ARC usage by tweaking the 'primarycache' setting on any ZFS or ZVOL | 17:51 |
| vlm | sdeziel: just random search on web says about 1G ram on 1TB of storage | 17:51 |
| sdeziel | on tightly constrained machines (512M of available RAM), I ran with just metadata caching | 17:51 |
| vlm | sdeziel: what size of pool? | 17:52 |
| === ijohnson is now known as ijohnson|lunch | ||
| sdeziel | vlm: it was a mirror of 2x 750GB | 18:30 |
| vlm | sdeziel: nice to be able to run zfs on such small memory,though saw someone mention something regarding writes and block size,so if turn off performance=all and just metadata it would result in much heavier write operations on disk? | 18:39 |
| sdeziel | vlm: the ARC is mostly (exclusively?) for reads | 18:40 |
| sdeziel | vlm: I'm not advising to tweak primarycache blindly. This should only be done if you feel the ARC doesn't shrink enough when there is memory pressure | 18:42 |
| vlm | sdeziel: considering adding some maybe,ill see how it turns out first | 18:43 |
| sdeziel | vlm: I would suggest you run with stock default params for a while and then only try to fine tune a thing or two. The ZFS folks have worked hard to get good defaults and they know their stuff way better than I do | 18:43 |
| sdeziel | vlm: ZFS will happily take all the RAM you throw at it but it can accommodate surprisingly well with very little too. | 18:44 |
| sdeziel | vlm: how much RAM do you have and how much data are we talking about? | 18:45 |
| sdeziel | BTW, the arc_summary script will show you a lot of nice stats if that interest you | 18:45 |
| vlm | sdeziel: im about 4GB though i got a lot going so think im free about 1-2GB | 18:46 |
| vlm | sdeziel: im off abit thanks for all the usefull information that script also very informative,ill do some reading on info zpool aswell whole other world this compared to mdadm -_- | 18:49 |
| sdeziel | vlm: you are welcome | 18:49 |
| === ijohnson|lunch is now known as ijohnson | ||
| sveinse | I'm running 18.04 server and I'm setting up ipv6 with SLAAC and I need the server to generate a stable IP address. I notice that my ubuntu 18.04 and 20.04 does this out of box, but the server does not. Where might I find more info about this? | 20:24 |
| sveinse | *my ubuntu 18.04 and 20.04 desktop | 20:25 |
| RoyK | sveinse: you can turn off ipv6 privacy extensions - that'll make the SLAAC address based on the MAC address | 20:59 |
| RoyK | sveinse: very useful if privacy isn't relevant | 20:59 |
| RoyK | sveinse: I used that for a bunch of infoscreens running off raspberry pi machines, the hostpart of their ip defining the machine and the webserver using that to serve the correct data | 21:00 |
| RoyK | easy peasy :D | 21:00 |
| sarnold | RoyK: ooo that sounds neat | 21:01 |
| RoyK | sarnold: scales well too ;) | 21:03 |
| sveinse | RoyK: my ubuntu desktop creates two ips, one "temporary dynamic" and one "dynamic mngtmpaddr". And I believe the latter is a stable IP but still using the priv extensions. This is what I want for my server too. | 21:04 |
| RoyK | sveinse: for the server, I'd use a static IP, not SLAAC, but then, that's perhaps just me (although I doubt it's "just" me) | 21:08 |
| RoyK | remember that if you choose a SLAAC address for a server and the server is replaced, it'll get a new IP. Just use a static address for those. It's easier unless you have a *lot* of servers to manage, in which case there are other ways to sort it out | 21:10 |
| sveinse | RoyK: yes. I want to skip setting up DHCPv6, and nor do I really want to set static IPs on each server. So I would have hoped SLAAC could save me the trouble, but I do indeed see the contradiction is what I want to achieve | 21:12 |
| RoyK | sveinse: may I ask what sort of environment this is? | 21:14 |
| sveinse | I wonder which is least troublesome DHCPv6 for central config of IP addresses or on each of the servers :D For ipv4 its central DHCP today | 21:15 |
| RoyK | (probably up north, so that dhcpv6 will freeze during winter) | 21:15 |
| sveinse | RoyK: A small office network. Around 20-ish servers with fixed ip | 21:16 |
| RoyK | ok | 21:16 |
| RoyK | then it might perhaps be just as much work setting up dhcpv6 than setting dhcpv6 addresses manually? ;) | 21:17 |
| sveinse | For some unexplained reason I have a mental impression that SLAAC is better than DHCPv6 due to it being stateless. Thus for most clients and users, I do want SLAAC for them. | 21:18 |
| RoyK | not as fancy, but if set manually, it'll work regardless of the dhcp server running or not | 21:18 |
| sveinse | The reason for having ipv4 DHCP assignment of static resources is because then the dhcp server is the one stop shop for assignments | 21:18 |
| RoyK | we still stick to static ip addresses on servers at work, and we have a wee bit more than 20 servers | 21:19 |
| sveinse | RoyK: yeah, another company with 1400 employees does that too on its servers | 21:20 |
| RoyK | sveinse: I work for oslomet | 21:20 |
| sveinse | RoyK: oh, cool, this is in Stavanger | 21:21 |
| RoyK | I guessed somewhere in .no by your name and ip :) | 21:21 |
| sveinse | May I ask if you use any of the link ips, fe80::, for anything for the network infrastructure? Or only the global scoped ips? | 21:23 |
| sveinse | I noticed that windows tend to use fe80:: as default GW, while linux use the global | 21:24 |
| RoyK | I'm not sure about that, as for default gateway or similar. there might be something out there. I haven't worked too much with the networking details | 21:24 |
| sveinse | ok, thanks | 21:24 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!