@ewwhite My home server runs ZFS just as well as OpenSolaris; except for the iSCSI and SMB integration crap that makes my neck hair stand on end anyway.
I worry more about the super "creative" solutions people are concocting.... Like the above. I can't think of any reason it wouldn't work, but it seems so cobbled together that it probably wont work, if that makes any sense.
@MDMarra Oh yeah; that stuff I completely agree with, and it was nice that you explained more than "no, that's wrong". Too bad people don't get rep for insightful comments.
I need a tool/script that can help me find local user accounts that do not have passwords across a network. Not all computers will be on the domain, but 90% are (win2003). But I need to audit/check local user accounts, not domain user accounts.
I am making an haproxy module, part of what it does is generate the config files. There is a lot of config, but it seems like init.pp is "the place" for manifest "data". Anyone have any recommendations for breaking that up?
I accidentally poured some mortar and grout down my bathroom drain when I was cleaning the remains of a bucket of mortar after tiling the kitchen. The drain was plugged up pretty bad so I tried to scrape some of it out with a wire hanger down the overflow, and that got it flowing again, but its s...
@lsiunsuex That reminds me... Most people think Cement dries, as in the water evaporates out and that makes it hard... So if you keep cement wet it will not dry. This is very wrong, cement is a chemical reaction catalyzed by water; once it gets wet it will harden (with only a few exceptions, none of which involve water).
i've been thinking about this one now for a bit - when i wake up, goto take a shower - as the shower water heats up i drink a cup or 2 of water from the sink tap to wake me up - how can you know if the water has been contaminated some how (thinking the drug they put in the water in batman begins) that early in the morning. if something happened over night that you slept through and have no idea whats in the water this morning
@lsiunsuex Well, unless you run your water for ages, you're still going to be using water that's well within your property boundary. Any kind of contaminant would have to be seriously powerful in tiny quantities to dissolve in that amount of water and have any effect
@Dan I know you're across the pond and all, but here our Counties have "Hazardous Disposal Information" available, that details places that will take all sorts of stuff. The vast majority are free too.
@WesleyDavid was just curious how living there was; i've been interested in moving to arizona / nevada for years. how cold winters are? bug situation? (snakes, shit like that)
well, I know that in my town you can dispose of every hazardous product to the same place, so it's much more easy to dispose of them. But don't know how it work were you live
Still some of my neighboors burn them in their yard (the police came often to fine them though, but they still do it)
@ewwhite I saw your earlier messages but I've been upgrading fabric switch firmware all morning. If they can't do replication properly, they won't get taken seriously by any big shops. That said, even big shops have unreplicated storage- one of our two core prod boxes is unreplicated (for pre-prod), and all of our remote locations have unreplicated storage.
@ewwhite Right, and unless they're stumbling over themselves trying to tell you how it works, how well it works, and how many other clients are using it to great effect, they're not pitching it properly
that said, storage replication isn't the only choice these days with virtualization
@lsiunsuex Winters are magnificent. I should say that I love hot weather, and even the last week's 115 F days didn't bother me, so I"m a bit unusual in that regard.
@Basil they didn't have good answers for that. So, should replication be a big deal in storage solutions? It sounds good in theory, but what do people actually DO with it?
It's dry heat. You'll hear people whine "But when its 115 it really doesn't matter" - that's total bullshit. 115 at single digit humidity is wonderful. 98F at 90% humidity is hard to deal with.
@ewwhite that comes down again to the whole enterprise versus mid-market question. What people do is reduce their insurance policy costs by testing recovery at a remote site, having a defined and executed recovery time and point
@Dan Depends on which type of brake fluid. DOT 3, 4, and 5 are very different beasties. I have no idea how they'd correlate to whatever labelling is used in the UK either.
It's not a direct answer to your question, but a more traditional architecture for this sort of thing would be to use HAST and CARP to take care of the storage redundancy.
A basic outline (see the linked documentation for better details):
Machine A ("Master")
Configure the HAST daemon &am...
"There is a right (or at least documented/supported) way to do this" :-P
@ewwhite Why wouldn't they? ZFS is a fine storage platform.
@Mami Aside from home networks being off-topic (SuperUser is for hope stuff) there is no possible way to answer this question with the information you have given. Asking "Why is login slow?" is like calling your doctor and saying "Why does my arm hurt?" - it could be idiopathic pain, or you could have got your arm caught in a combine wheat thresher. There's no way to know & we won't guess. You are the one with access to the system, so you must perform some kind of troubleshooting and provide enough information to help other people come to a diagnosis. — voretaq71 min ago
@ewwhite Local storage isn't hard :) Their big problem is that they're competing with the "roll it yourself" folks who have already decided not to go the enterprise route.
@basil they added AFP (why?), and add their tuning, added call-home and designed their JBOD enclosures... manufactured by the same provider that supplies NetApp and Sun enclosures.
"The manager wants to bring you in for an in person interview this Friday. The email bellow was send to you buy mistake. Please let me know if you are available. "
Basically what I am trying to accomplish is having Server2008 (guest) run on my Win7(host) with another Win7 (guest) as well as a WinXP(guest) on a mac book pro(host)...
I was pointed in the UDP Tunnel section but I just do not understand their command line.
This is how my network is layed out
...
@voretaq7: thx for hints rgd serverfault.com/questions/419679/… / I generally agree: better use something proven .. but there are upsides with the proposed stuff - if it could be made working reliably
@oberstet Like I said, what your talking about can definitely be done - I've just never seen a configuration like it. My experience with iSCSI is that it's not quite as reliable as SANs (switch failures, errant idiots unplugging cables), so not knowing how it would work in practice I'd be wary of implementing it without going through lots of testing
@ewwhite why you flags things? I'm standing right here :P
@voretaq7: I have seen ZFS mirror/raidz over SAN LUNs (which likely had redundancy already/anyway) .. but you say: the thing unusual is to have that with an iSCSI based SAN?
@oberstet Also, iSCSI doesn't handle well when you're streaming lots of IO to it. Might be worth considering SAS
iSCSI has no flow control, so can and will drop packets, which leads to out-of-order delivery on scsi commands. This will cause weird failures and latency that are no fault of the disk
@Basil: how do you "remote" SAS over Ethernet? is that possible? even when: that then requires a switched Ethernet only, cannot cross routers. Doesnt iSCSI use TCP and hence has flow-control?
@oberstet tcp doesn't cut it for SCSI. You need FCoE if you want the kind of flow control that SCSI needs.
@oberstet TCP's flow control is to drop packets without informing the sender when they get full. That is supposed to be handled by the underlying protocol, but in SCSI's case, it doesn't handle it.
This means a read or a write might just get dropped and not resent for seconds, while all the other ones in the burst were done in, whatever, 6 ms
it's not a problem if your bandwidth is greater than the maximum performance of the disks, though
@oberstet If you had a FC SAN, you wouldn't need ZFS. All the RAID etc is handled by the storage controllers, probably more efficiently and at a higher performance than a server can do it
@Basil: I trust ZFS a lot more than any SAN RAID and especially HW RAID controllers. And in any case, having a "big" SAN storage center doesn't fit with my main motivation: being able to use wimpy nodes that will fail, but are cheap and can scale out.
@oberstet You can probably make it work with iSCSI, but I wouldn't trust it without stress testing (and like @Basil mentioned I'd include huge IO / link saturation to see what happens to performance. Database servers can be IO heavy and you don't want an unpleasant surprise...)
@oberstet then you're going to have to live with the performance limitations of iSCSI. Unless you can find some way of using FC or FCoE to do the underlying disk IO.
@oberstet your trust is sorely misplaced. I would trust an enterprise SAN vendor's RAID implementation (HW RAID 1, 10, or 6) over ZFS -- these guys have been doing the RAID thing for a LONG time, and they have to get it right: Storage isn't allowed to fail :)
Not that ZFS ZPool/RaidZ is unreliable - quite the opposite actually I've never heard of a failure in production - but there's no reason not to trust (e.g.) EMC to get storage right
@oberstet True, but people use EBS because Amazon has gotten the right balance of "mostly reliable and dirt fucking cheap" for them :-) Honestly if you implement the ZFS design you're talking about or the HAST+CARP design I mentioned you'll still probably do better than EBS (provided you test the hell out of it before you deploy it)
@Basil, @voretaq7: Ok, lets put it like this: however you look at EMC/HDS based stuff (maybe the most reliable in the world, maybe not): it is absolutely expensive stuff. And it does not run on a bunch of wimpy PC servers. And that is what I am targeting ...
My concern with your ZFS design isn't so much "Will it work" -- of course it will, it's (relatively) standard ZFS features -- It's "How will it handle when things break?"
You don't want to be an Amazon where you have a switch fry or a node blow up and all of a sudden your environment goes all catawampus
@oberstet yup - which puts you in the territory of "use a proven solution" or "prove a new (possibly better) solution" :-)
With EBS, people often layer SW RAID on top of LUNs .. to "improve" performance (haha) or for additional "security" (maybe). "prove a new solution": I agree. Instead of babbling I should do that. Ha. Will be a heck lot of work ..
@voretaq7: Having unlimited budget comes nifty - only had that when working for a telco .. at that time they ran on Oracle RAC, HP Superdome and HDS. Don't even ask for the price tags. We don't have unlimited budget.
I don't really know if Server Fault is the idea place to be posting this, but I'm more interested in people's opinions than anything else.
We currently outsource a datacenter to a 3rd party who, for the last 5 years, has provided us impeccible service at a really good price. In the last 12 month...
@MDMarra Meh. I have LDAP-centralized authentication already - I see no need to pay Microsoft a license fee and buy 2 more servers just to use AD in my otherwise all-unix environment :-)
@voretaq7 Going to be using NIS to import into AD and then force our Devs to have the in-house app update straight into AD via LDAP hooks
My new Padawan is cool with the Windows AD stuff from her courses in tech school, just a matter of getting her an opportunity to shine. She's good, but she's not very confident.
@voretaq7 Yep. Boss is even starting to get interested in this project too since it will take care of our unencrypted password problem which would likely block our EHR certification this spring.
Or at least that's what it sounds like he's saying
@Iain yeah a lot of those we probably want to pass through in chat first (at a quick glance ports is one that may have some funkyness and need to be split into TCP/UDP ports versus [tag:$OS-ports]
I know the formatting isn't fantastic, but is the question really unclear as it is? (e.g. I put through your edits to the answer & made the comment because they illustrated how to use the referenced tools - which is important and was missing from the original answer)
@voretaq7 Oh no, the question is fine, I guess. Personally, I especially like to clean up our old questions so they can serve as examples to new users.
@OliverSalzburg not unless they're egregiously bad or have offensive grammar (or if they're on our bad tags list and I'm in there for other reasons)
@OliverSalzburg If you see really awful ones with munged formatting (like no code blocks or something) or 5-page signatures by all means send the edit through - those I usually approve :-)
@OliverSalzburg Also I freely admit edit approval from me is a like a pachinko machine - what I reject on monday I might approve on wednesday depending on how cranky I feel :)
So, having read Kyle's linky to the Puppet blog, to be able to use Puppet effecitively, I need to go learn Ruby, YAML, and JSON. Yeah, like I'll have time to go do that anytime soon.
I need a tool/script that can help me find local user accounts that do not have passwords across a network. Not all computers will be on the domain, but 90% are (win2003). But I need to audit/check local user accounts, not domain user accounts.
It's not a direct answer to your question, but a more traditional architecture for this sort of thing would be to use HAST and CARP to take care of the storage redundancy.
A basic outline (see the linked documentation for better details):
Machine A ("Master")
Configure the HAST daemon &am...
