| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Comments:
<0> Really...sometimes it's so hard to keep your cool when dealing with external suppliers
<0> Luckily it wasn't really my problem, but I get worked up over it anyway
<0> If only they'd see the light
<1> PolarWolf: read a very nice rant in asr the other day - from a guy who was pig sick of fighting to get ios updates from cisco, replaced a dozen of them with linux boxes running the quagga suite, and noticed they were more efficient and configurable for a quarter the cost...
<0> DaveHowe: Yeah. Though I guess a true cisco has a nice and fast backplane to work with, while a PC is stuck with PCI{e,-X)
<0> But hey...as long as it keeps up with line speed, who cares
<0> Hell, cisco sells PCs^WPIX too, so I guess it's good enough
<1> PolarWolf: pretty much, although tbh it would take more than 3 gigE cards to saturate a pcix bus, ***uming you have only one controller
<0> DaveHowe: Yeah. And with TOE these days...
<1> cpu tends to be the bottleneck - and the cpu in most ciscos ****s
<1> although (again) on a cisco, it doesn't have to worry about the fastpath packets
<1> card-to-card fastpathing on a pc is unusual - it can be done, but the cards end up costing more than the cisco would
<0> DaveHowe: OTOH, for one crisco you can add a ****load of PC computing power
<1> PolarWolf: yeah, plus if you need to debug a cisco, you have to force *everything* up into routing space, which means the cpu ends up on its knees begging for mercy
<0> Anyway...it was nice ranting
<0> Gotta get some sleep
<0> DaveHowe: Hehe
<0> DaveHowe: I'm not a networking guy, s I wouldn't know :)
<0> I just know bits and pieces, enough to get by
<1> PolarWolf: be glad. I have been forced to deal with the insanity that is N3 (Connecting For Health)
<0> And it's more than the network admins I have to work with do
<0> "I need a port in the production VLAN"
<1> BT got the contract, and have seriously underspecced the pixen they use for last-hop firewalling of sites
<0> "Which VLAN number is that"
<0> "Well WTF, that's your ****ing job to figure out"
<0> "Twat"
<0> Too bad I can't use profanity at work
<1> to the extent that a lot of them rely on the bloody *routing tables* to keep Bad Packets away...
<2> PolarWolf, have any experience with RAID arrays ?
<0> Well, not too explicit anyway
<0> fbset: Some
<1> Redundant arrays of inexpensive druids?
<2> PolarWolf, can you help me with a bit of a RAID problem please ?
<0> fbset: Though I try to make it easy on myself and buy hardware controller :)
<2> ok here it goes then
<0> fbset: No, I'm gonna get to sleep
<3> PolarWolf: heh, I deal with soldiers day in and day out.
<3> PolarWolf: so profanity is expected. ;)
<1> PolarWolf: next rant: the gui "support set" HP supply for their raid hardware for linux, that is for the 2.4 kernel only...
<0> Strider: Lucky sod :)
<4> tthat reminds me. i wonder if tojoe ever figured out his pix ipsec issue
<2> PolarWolf, just a quick idea
<3> my favourite line of one person describing another "He may be a Lieutenant Colonel in the army, but he talks like an admeral"
<1> PolarWolf: despite the Vista "support set" being out already
<0> DaveHowe: Dealing with HP is...
<0> Well
<0> Like trying to kill vampires
<1> PolarWolf: their hardware is popular, unfortunately
<1> PolarWolf: and the command line tools work fine
<0> DaveHowe: Yesh. And they'll promise you the world, and don't deliver
<1> even if the command syntax is a bit stilted
<0> Hah, I have another rant...
<0> No, **** it...sleep time :)
<2> PolarWolf, i have a server that had the raid broken.. 3x120Gb hdd totally and i replaced one faulty drive(samsung) with an new one wd...
<4> im going to let my mind go numb while watching tv.
<1> "hpactrl controller slot=0 array a volume 0 add hotspare drive 4" sorta crap
<2> PolarWolf, the new one added that is 3 cylinders short, so i need to downsize hdb2 from 274 to 271 and have SWAP parition raided between the two drives
<5> D-side: and this is different from any other time... how?
<0> fbset: Making partitions smaller...bad, bad, bad idea
<2> PolarWolf, http://pastebin.ca/343320 ,
<2> PolarWolf, well what to do then ?
<1> fbset: get a bigger disk
<0> Really, get a drive which matches and save yourself the pain
<0> Bigger works
<1> fbset: do not try to shrink a stripe set which is already in failover mode
<2> i know md should detect that one drive is bigger, and just use what it can off both
<1> fbset: or Bad Things will Happen
<2> but it doesn's
<1> no, it will use a section of the bigger drive which is equal in size to the area of the old drive
<2> PolarWolf, DaveHowe, 3 cylinder short oh come on and yes i know: a -very- good reason to always leave a few slack cylinders at the end of a drive in a raid
<2> but it should work with resizing
<6> i i have a question i'm studing network engeneering and i like to use linux in my project wich consesting of a migration from ipv4 tp IPv6 and i like to use a better distribution for this job any one has an idea?
<0> It's your filesystem you're gonna **** up. I'm not going to recommend you even going there.
<0> Bigger...great. Smaller; never.
<1> fbset: Bad Thing has been ***erted. if you want to go ahead and do it anyhow, fine, but that's your data.
<1> arpf: anything 2.6 based should be fine. in fact, any distro updated in the last year should be fine :)
<6> good DaveHowe !!
<3> heh..
<3> i've pondered rebalancing a raid, under lvm2 under ext3.. on a live system.
<0> pw@shanta:~$ ip -6 r | wc -l
<0> 16
<3> or is that just playing with thermonuclear weapons?
<0> Whee!
<2> # ip -6 r | wc -l
<2> 7
<1> Strider: I have pondered the effects of diving a fully laiden 747 into a nuclear reactor. Doesn't mean I have any desire to see it happen :)
<0> Man, playing with your data on storage levels is...ballsy :)
<1> Strider: anyhow - LVM2? not EVMS? :)
<0> Though sometimes necessary
<3> PolarWolf: LVM2 does make it fairly safe, imho.
<3> but i'm a tad leary about re-aligning a RAID 5 array. :D
<0> Strider: Yeah, true
<3> well, re-striping.
<1> PolarWolf: RAID is too crude and low level to do stuff like that safely. LVM{2] makes it noticably safer, and EVMS makes it almost routine :)
<0> DaveHowe: These days I just ask my SAN guys for an extra LUN :)
<2> DaveHowe, PolarWolf but judging from the data provided here http://pastebin.ca/343320 it really matters that much? , if the differences are quite very small 3 cyl ?
<1> PolarWolf: that said, most raid hardware can safely increase the stripeset
<0> fbset: It's your risk
<2> PolarWolf, ok but you know how this is done ? :)
<0> fbset: resize filesystem, resize partition, resize volume set
<2> with fdiks?
<1> fbset: odds are good that each stripeset across the drives contains a continous section of virtual cylinders
<2> fdisk
<0> fbset: resizefs, parted and whichever tool you need for operating your RAID
<1> fbset: so in that case, you would lose a few cylinders from the end of the last partition
<2> yesd
<2> i need to downsize hdb2 from 274 to 271
<0> It's very likely your first step will already hose the filesystem. Provided resizefs won't give some really nasty errors in the firstplace
<1> PolarWolf: probably safe enough if he deliberately hoses the last partition first, resizes the stripeset, reboots so linux can "see" the new virtual volume size, then rebuilds the last partition
<0> I hope you keep decent backups
<1> PolarWolf: but I wouldn't want to have to do it :)
<2> it;s a Centos 4 with software raid
<1> software raid ****s, and is a horrible alternative to lvm
<7> fbset: can you say backup? come on now. it's B-A-C-K-U-P
<2> libolt, yes it;s all backed up
<0> DaveHowe: It *might* work, ***uming things aren't in a single partition
<7> good boy
<2> yep :)
<0> Oh, you have backups?
<2> yes
<0> Scrap the raid altogether, create a new one and restore
<0> There, done.
<1> PolarWolf: If the last partition is critical, I would try to shrink the partition first, then resize the filesystem while in single-user mode
<2> i knew/know it might come in handy
<7> heh I actually run two software raid 1 mirrors on my home file server with LVM on top of them
<1> PolarWolf: THEN resize the array and rebuild it with the extra drive in it
<6> another question which is supporting better the migration from ipV4 to IPv6 suse or fedora or mandriva or debian?
<1> libolt: that's an interesting example of your insanity. do you want to explain further?
<0> DaveHowe: Hmm, I take it you mean you wanna resize the filesystem first
<0> DaveHowe: I can't start to imagine what would happen if you do the partition first and then the filesystem
<1> arpf: there is no "migration" therefore the question is meaningless
<8> my connection keeps dropping
<7> DaveHowe: it's a home server, nothing extremely important :) Plus I've had no issues with software raid thus far, drive dies, toss a new one in, add it with mdadm and rebuild :)
<6> there is 2 methods to migrate routers from ipv4 to ipv6
<0> Actually, he might get away with it if he somehow defragments the filesystem and makes sure all the data is at the start of the partition
<1> PolarWolf: partition resize won't be noticed by the os until you reboot again
<0> libolt: Works for me too
<9> arpf: Does it really matter unless your ISP is an IP6 broker?
<3> PolarWolf: though unlikely.. bad **** happens if you don't first resize the filesystem
<1> libolt: not that. you are adding two redundancy abstraction layers (lvm and soft raid) where one (lvm) could quite happily do as many redundant copies as you want
<6> atompngn my project is about that
<6> i don't care for the ISP
<0> DaveHowe: Hm
<3> and that will happen the day pigs fly.
<2> PolarWolf, but why do i need backup if we are only messing with one small partition that is SWAP i do not need to touch ANYTHING but the swap partition
<2> ?
Return to
#linux
or
Go to some related
logs:
#c++
#mirc
ryan_18
sikmekistiyorum
#linux
#MissKitten
#linuxhelp
phone call
WHAT IS THE FAMILIAR NAME OF THE FAIR CITY ALSO KNOWN AS BAILE ATHA CLIATH ?
pegoeut
|
|