The Comms Room

kce

00:45

0

Q: How to configure multiple iSCSI Portal Groups on a EqualLogic PS6100?

I am working on a migration from a VMware vSphere environment to a Hyper-V Cluster utilizing Windows Server 2012 R2. The setup is pretty small, an EqualLogic PS6100e and two Dell PowerConnect 5424 switches and handful of R710s and R620s. The SAN was configured as a non-RFC1918 network that is not...

@MarkHenderson - Not sure if that is coherent but give it a look see if you are so inclined. I think I'm going to start a support case tomorrow.

Anonymous

@BigHomie lol you still hold that belief where I'm actually 21 and a supressed perv?

BigHomie

@PatoSáinz 35- 41 actually, but only sometimes

Anonymous

@BigHomie "sometimes"?

Anonymous

what do you think when you do not think that

Anonymous

01:00

truth is i'm Benjamin Button

Anonymous

►

Anonymous

for anybody that needs to train a user into the "microsoft suite"

Journeyman Geek

01:39

@PatoSáinz naw, you're clearly only 12.

Anonymous

01:58

@JourneymanGeek near but no

Chris S

03:20

Watching news - guy gets pulled over and blows twice the legal limit, admits to running over a stop sign. =]

Mark Henderson

@ChrisS Hah. I got breath tested twice on the way home yesterday - once just outside my work and once just outside my home

Just as well i didn't have five lunchtime scotches

EEAA

04:26

@MarkHenderson - I retracted my close votes on Tom's two questions. Still, we ought to ask ourselves, if those questions were being asked by anyone other than a sysadmin rockstar and SE employee, would they not certainly be closed due to lack of minimal understanding?

Mark Henderson

@EEAA I considered the same thing. Borderline I think.

I just assumed they were voted as OT

Once I saw your comment I did consider what you'd written

EEAA

Well, the "minimal understanding" is under OT, which is crappy IMHO

I'd like to think we don't give special treatment, but...

Mark Henderson

@EEAA Of course we do

We tolerate TomTom's tomfoolery. We'd have shitcanned anyone who normally talked like that if he wasn't always so right

But official policy is always "vote based on content, not person" so...

EEAA

It would all be so simple if it weren't for these dang humans.

Iain

05:46

G'day

dawud

06:24

morning, northerns

Jenny D

Morning

faker

06:46

Morning

Iain

07:37

Is there a moderator in the house ?

Tim Post

08:11

@Iain What's up?

Iain

@TimPost Hi thanks for dropping in

You can see it all here serverfault.com/questions/601112/…

The OP is a bit upset and needs calming down

Dennis Kaarsemaker

@JourneymanGeek poke

do you know any decent VPS providers around you?

Dan

@ewwhite You're going to have to move to GMT/BST I'm afraid

@Iain *killing

strugee

@Iain you should tell him to try Expert's Exchange

Iain

@Dan I'm being pink and fluffy

3

Dan

08:25

@Iain It doesn't suit, Mr Vader

Iain

:(

Dennis Kaarsemaker

08:35

pauska

g'day

RobM

so it's the 25 year anniversary of the tianemen square massacre of all those students today.

morning @pauska

Dan

@RobM And now China is everyones buddy

Unless you speak out against them, in which case good luck

Tom O'Connor

08:55

                            1111   111    11    111   1111   111    11    111
      544444454455444444544500004440004449009444000444000044400044590094440004
      466575857522658586255400005760009769009667000665000098600068090096560006
  100                       ****   ***   ****   ***   ****   ***   ****   **
   90                       ****   ***   ****   ***   ****   ***   ****   **
   80                       *##*   *#*   *##*   *#*   *##*   *#*   *##*   *#
   70                       *##*   ###   *##*   ###   *##*   ###   *##*   ##

Chopper3

@Dan China is SO fucked right now, they can't service their debts and will be in a very similar place to Greece by ~December. Oddly it'll actually benefit the US the most as the Chinese will have to sell off their US assets quickly and at a loss. Most of Western Europe will do quite well out of it too.

Tom O'Connor

Guess when I turned IPAM off. It was pinging every IP in RFC1918 every 10 seconds. This was causing the router's sw-forwarding queue to get improbably huge.

Iain

@TomO'Connor improbably or actually ?

Tom O'Connor

@Iain improbably. I'd have thought it would probably crash, rather than just get bigger and bigger.

Iain

:)

Tom O'Connor

09:05

it was about 200 billion packets when stopped ipam (actually disabled the VM's nic)

RobM

@Chopper3 so their bad economy will help ours improve. That's good... well good for us. I think another slump here would be murder.

pauska

09:31

@TomO'Connor what kind of IPAM is this? MS?

Tom O'Connor

09:45

@pauska SolarseholeWinds.

pauska

@TomO'Connor so if you have 10.000 RFC1918 adresses active, it would ping each one every 10 seconds?

Tom O'Connor

@pauska Something like that. I haven't really delved into it to figure out what it does and why

but it was pinging a shitload of addresses, some of which don't exist, so the router was going nuts trying to route them

bhuvin

Hi guys !

Just wanted your inputs :

Tom O'Connor

@bhuvin PS/2 for me.

@bhuvin Is that like a subtle sysadmin pick-up line?

bhuvin

I have multiple Instances of Servers having different modules ... Now these modules actually need to share some information amidst themselves ... The modules are capable of having TCP connections ... Then in that case What should be used ? Common Database or Server to Server TCP based communications ?

Tom O'Connor

09:57

@bhuvin modules of what?

bhuvin

@TomO'Connor - hahahaha ! Man that was a subtle "HELP ME" line ...

Tom O'Connor

You mean modular services?

pauska

I think you're in the wrong room mate

this is the sysadmin lounge, not programmers

Tom O'Connor

They should communicate over HTTP, or a message bus (RabbitMQ or similar AMQP broker)

bhuvin

Modules of my code : You can think about them as Separate TCP Servers ...

Tom O'Connor

09:58

@pauska And it's a sysadmin/sysarch question. If you let programmers make decisions like this, you end up with shit networks.

@bhuvin Common database incurs a single point of failure on the db service.

bhuvin

@TomO'Connor - Yeahh ! but those are the ways ...

Tom O'Connor

You want the loosest coupling possible.

You might be able to do it with Zookeeper..

depends.

what they're sharing.

Redis might be a reasonable fit, too.. in a strange way.

bhuvin

Actually the purpose is like the Customers are connected to the TCP server , Now since there are Load balancers in between , So the Customer might be connected to Any of the servers in the grid

Tom O'Connor

I mean.. There's a whole bunch of techs you could use.

Y'know.. It probably doesn't matter that much. what have you tried so far?

bhuvin

But now there is a requirement :

C1 is connected to S1 and C2 is Connected to S2 . C1 needs to Send a Message to C2 , Now in this case S2 needs to forward that message to S1.

Tom O'Connor

10:03

See, if the servers sat on a message bus, they could both receive it..

How about multicast?

So they all get the message..

bhuvin

Like how would you share a Message bus ?

I mean there has to be a Common place where they could share the state ,,,

Yeah ... Sort of that is another way to do it ...

but i wanted to know ...

Tom O'Connor

C1 injects a message. S1 receives it, but doesn't remove it, S2 receives it, Sn receives it and removes it

bhuvin

Conceptually why it should or Shouldnt be done ?

I mean Server to Server communication ... ?

Tom O'Connor

but you'd need some way to ensure that messages get removed once they've been actually processed by all servers.

@bhuvin There's a deep question.

bhuvin

Yeah ... Dats into place ...

Tom O'Connor

10:05

But you'd have the message removal problem with a database, too.

bhuvin

@TomO'Connor - dats wt i needto know and take a call on ...

Whether we should avoid server to server communication in this way or not ? Actually speaking its not that that these Servers are different servers but just the Replicas of each other ...

Tom O'Connor

@bhuvin It's one hell of a double edged sword. Are they maintaining state indivudually, or connected to a central database cluster?

i mean, to store the shit the customer sends in

There are no right answers, incidentally.

bhuvin

Central db store ..

@TomO'Connor - Exactly ... Its like more of a Moral science question ...

:)

Tom O'Connor

Your best bet, is to figure out a bunch of options:
Database.
Message Queue (Publish/Subscribe)
Multicast
Peer-to-Peer connection

Then try each one, and see what works best for you.

There'd be something incredibly elegant about actually finding a use for Multicast after all these years.

bhuvin

@TomO'Connor - Working current yes ...

haha ...

Multicast i can handle ..

But the thing is, Is Server - Server communication in this way a good thing to do ?

Tom O'Connor

10:10

@bhuvin If it works for you.

As in, if it solves your business problems, then yes.

Doing it right is a different matter.. but You are Not Google. (Unless you are, in which case I'm wondering why you don't ask someone else)

So by the time you need to scale the service, you'll have figured out what doesn't work for you, and why.

bhuvin

haha ..

@TomO'Connor - Doing it the Right way is important ... Coz Yes i am not google but Google wasnt Google the day it started ...

Tom O'Connor

@bhuvin And they did stuff wrong at first too.

Which is why it doesn't matter, really.

bhuvin

@TomO'Connor - True ... But Delibrately due to Carelessness doing it isnt a thing to do either .. :)

@TomO'Connor - Like what could be forseeable issues ?

Tom O'Connor

@bhuvin Not knowing your product, that's also unanswerable.

@bhuvin It's not deliberate, they're unknown unknowns.

You can't mitigate problems you haven't encountered yet.

To some extent, it's premature optimisation (as embarassing as premature ejaculation, although easier to fix.) Just stop fiddling with it.

Jenny D

@bhuvin You may want to read up on the CAP theorem. You can have any two of Consistency, Availability and Partition-tolerance. Making up your mind about which one is the least important for you early will save a lot of headache later on.

Tom O'Connor

10:19

@JennyD You can kind of have 3 if you'll settle for eventual consistency.

but nobody ever does.

Jenny D

@TomO'Connor yeah - but again, then you've made up your mind that concistency need tnot be 100% for 100% of the time. Knowing in advance that you have to make that choice will clarify a lot of things.

Tom O'Connor

Most tend to prefer C&A, and forget about P.

Jenny D

@TomO'Connor My cynical self says that most don't even know about the issues and once they do they've built themselves into a corner.

Tom O'Connor

@JennyD The consultancy trough. Companies are productive until they reach that, then they need to hire someone with a mop and bucket.

Jenny D

@TomO'Connor ...which is where I make my money, so I shouldn't complain :-)

bhuvin

10:24

@JennyD- Surely will go through the same ... Currently in this scenario, I guess it doesnt make a lot of difference in case of either approach, but isn't it a Norm that such Replica - Server shouldn't communicate in this way ?

Tom O'Connor

@JennyD Same here.

I'd like to see more systems using single system image.. like true clustering.

bhuvin

@TomO'Connor , @JennyD - having your own startup doesnt facilitate the same , So you are the one with Bucket and the Mop ... So wanting to avoid the same ... :)

Jenny D

@bhuvin There are many different ways to do this. I've been in a place where availability and partition-tolerance were essential - at that place, we used UDP to broadcast information between servers and each server updated their internal SHM hen they got a packet, and issued their own packets when they needed to.

Tom O'Connor

@JennyD I think that'd be supersexy over Multicast, rather than udp broadcast.

Jenny D

I've also been in a place where all three were important but we could skimp on consistency for short periods of time - we used Erlang for data sharing.

Tom O'Connor

10:27

Thing is, there really are no right answers, really.

because it's all a bit too personal.

If you really want this problem solved, pay either JennyD or me, a bucketload of money, and we can.

Jenny D

@TomO'Connor I think it was multicast, actually. Been a while, I don't recall all the details.

Tom O'Connor

But not without getting inside your business model with thigh-high waders and those long veterinary gloves.

Jenny D

And I've been in a place where consistency was paramount; at that place every server used the same database and if they lost touch with it they stopped handling clients.

Like @TomO'Connor says, you need to actually spend the time and money to get this right from the start. If you can't afford it as a startup, then you'll at least need to build something that is modular enough that you can extend/replace the communications bit later on.

Tom O'Connor

@JennyD I'd probably go down the AMQP route here.

Probably.

I dunno.

Jenny D

@TomO'Connor Yeah. But it has its own problems too, of course.

But then, everything does :-)

Tom O'Connor

10:35

@JennyD Indeed.

it does give you a lot more in terms of message validity, checking, acknowledgement, removal

which you'd have to craft yourself going down the udp route.

It'd almost be worth finding a protocol agnostic message broker

Jenny D

@TomO'Connor yeah - the place that went the udp route already had some of that setup for use via unix socket, plugging in a udp listener wasn't hard. And since consistency wasn't a priority, there was no ack or removal - they would either get the message or they wouldn't, no time spent on checking which.

It worked for that specific purpose but it's not a route I'd generally want to go.

Dennis Kaarsemaker

►

trying to de-monday-ize today

faker

one of our new products has "blacklist" in it's name. Now people are scared that sending mails from a domain with blacklist in it could cause e-mails being dropped.
Is that a valid concern? I'm not that much into e-mail/mass-mailing

Instead they've setup a new domain, screwed up SPF records and are even more wondering why mails don't arrive - but that's a different problem :D

pauska

10:51

Either you do SPF 100% correct, or you don't do SPF..

Dennis Kaarsemaker

@faker unless the domain is blacklist.me, I wouldn't worry to much :)

faker

@DennisKaarsemaker well, uhm. This is awkward...

Oliver Salzburg

When I deploy a MSI through a GPO and then manually uninstall the software on a client PC, should the package be reinstalled upon rebooting the system?

I assumed that it would be reinstalled, but it seems like it isn't and it's driving me crazy :P

Tom O'Connor

@OliverSalzburg Is it in Replace mode?

Oliver Salzburg

@TomO'Connor I don't really know what that means actually

And this system is localized to German, so that's an additional issue with making sense of it :P

> Yes, this is default and working correctly. AppMgmt remembers the GPOs and the assigned MSI packages that it already processed, and it will not process them again until you redeploy the package (in GPO editor, right click "all tasks"...)

From social.technet.microsoft.com/Forums/windowsserver/en-US/…

That seems to answer my question

pauska

11:20

deploying msi's through gpo is the most evil thing MS every invented

I was in the same boat as you about 5 years ago, and I've never ever used it after that..

3

Q: Redeploying Office to a single client after manual uninstall

One of our client PC's recently got random crash issues with MS Word. I manually uninstalled Office 2003 locally on the client, ran "gpupdate /force /boot" and expected the client to reinstall Office after rebooting. It didn't. I've ran gpupdate a couple of times, but nothing seems to take effe...

active-directory group-policy deployment ms-office

there is a solution there as well, if you don't want to redeploy the MSI to every single computer that the gpo targets..

Oliver Salzburg

Seems like it. I'm currently trying to determine how to properly reinstall the package on my test client without reinstalling it on all clients

@pauska Oh, awesome

pauska

My best advice is to never ever use software distribution in GPO's

buy SCCM.. or script it.. anything but gpo

Oliver Salzburg

I'm currently trying to publish the package in addition to assigning it so that I could reinstall it through the control panel, but the publish option is greyed out :P

Ah, it only works for users, not computers

I gave up and reinstalled the computer. Case closed. — pauska Nov 12 '09 at 6:47

Heheh

pauska

:(

Michael Hampton

11:46

> location Scotland, United Kingdom

Hm, someone's voting against secession...

pauska

11:58

@MarkHenderson Your bosses will love the extra licensing cost :(

Nimble Storage does not currently support storing application data in virtual hard drives (neither VHD nor VHDX) if the data requires a quiesce prior to snapshot. This includes database and transaction logs for applications such as SQL Server and Exchange.

wat

BigHomie

@pauska Wow

pauska

I don't get it.. is this the same for all arrays? Shouldn't Hyper-V pass on the VSS request to a SAN?

BigHomie

@pauska Over my head. Are you saying b/c your Hyper-V server stores it's VHDs on a SAN drive? I don't think that would matter, but I could be wrong

pauska

Well, any other SAN vendor I've talked to provide VSS agents for Hyper-V

BigHomie

ahh, I see now

pauska

12:09

connect.nimblestorage.com/thread/1362

BigHomie

What version of Hyper-V are you running?

pauska

12:24

Running? Well, WS 2012 R2 in our lab.. our production environment is vSphere 5.1 against EMC VNX

BigHomie

thinking of switching? Or just poking around?

ewwhite

Magellan

@pauska good to know re: VHDs and Nimble. Thanks.

Dan

12:40

@ewwhite I replied to your mail

Here's a question

If I snapshot a VM, then gracefully disjoin it from the domain. Can I roll back the snapshot and will the domain link be valid?

Nathan C

@Dan No, the computer object would be gone from AD

Dan

@NathanC No?

Nathan C

Whenever I've removed a machine from our domain, the computer object goes away too

Dan

not for me

Is that a version thing

jscott

Not gone, but disabled. Unjoining a computer just disables the object.

Dan

12:45

@jscott Sorry, I should have said after re-enabling, too

Nathan C

Huh, probably just ADUC being weird

jscott

@NathanC Well if you delete from ADUC, then sure it's gone. But if you unjoin from the client OS end it will disable the AD object.

@Dan Hmmm, never tested. I wonder if it would work, as the snapshot should still have the valid machine password. Please let us know if you test it, I would be interested to know.

Dan

@jscott I'll test it later for you

Nathan C

Huh...trying to enable AD recycle bin but I keep getting "insufficient access rights" ...prompt is run as administrator, i have enterprise admin

ewwhite

@Dan I see. I can't do anything until the client gets onsite.

Nathan C

12:59

Got it with ldp.exe...

Dan

@ewwhite Yeah, it's a ballache with stuff like this. Is it a major headache for you to be on site?

ewwhite

@Dan 1,700 miles.

@Dan The thin clients worked before and I could VNC to them

now they seem bricked

Dan

@ewwhite Hard to say without seeing, but the good news is the recovery is as simple as downloading the image from HP Support and using the util to put it on a pendrive

What did you do to them?

ewwhite

@Dan Nothing

one seems to have never worked

Dan

@NathanC @jscott Reverting the snapshot worked fine, tested logging on with brand new user account

ewwhite

13:13

client connected them headless to the network

so I could get a few to practice with

Dan

@ewwhite Oh, you really need to see what's on screen if they're not talking

ewwhite

Yes, a Windows STOP error

but another question... why would one capture/deploy an image instead of just copying settings?

jscott

@Dan Cool, thanks. That's useful to know.

Iain

wonders how accurate this prediction will be express.co.uk/news/nature/480038/…

HopelessN00b

@NathanC What's your functional levels? IIRC, not having a high enough functional level reports as a permissions error.

pauska

13:18

@ewwhite it's easier because of the write filter

(they run cold freeze)

BigHomie

@Iain Keep us posted

ewwhite

@pauska Well, I'm having a hard time with the HPDM and the handful of thin clients I've tried to deploy

HopelessN00b

@BigHomie Well, unless he gets killed by a tornado or a hailstone.

BigHomie

@HopelessN00b Then his absence will indeed be keeping us posted

Sorry to see you go @Iain, it's been a pleasure

haha, perhaps your oneliner works as intended, and there are no disabled users outside the disabled OU? — BigHomie 9 secs ago

pauska

@ewwhite never used it to be honest.. the only thing I can say is that imaging of the WE models takes a long time

Nathan C

13:23

@HopelessN00b 2008R2 for both

I was able to enable it with ldp.exe

BigHomie

@MichaelHampton so this means no brazzers.com placeholders here? I don't use it but was growing quite fond of eng.mit.edu

HopelessN00b

@NathanC Hmm, weird. I'd blame the previous admin and move on to the next fire/oddity/bullshit request.

ewwhite

@pauska All I need to replicate are connection settings

Michael Hampton

@BigHomie No, those are fine. It's the ones like mydomain.com which are seedy little sites we really shouldn't be sending anyone to.

Dan

@ewwhite Just deploy images

Way easier

ewwhite

13:26

@Dan clearly not... as the thin clients can no longer be contacted.

BigHomie

@MichaelHampton Whoa, speaking of which, would it be unethical to install Drupal on MIT's webserver?

tsss, pwned

Michael Hampton

@BigHomie Damn.

BigHomie

ikr

Dan

@ewwhite Oh, well yeah, but that's just not right

Nathan C

ERP system is burning...awesome

gets fire extinguisher

ewwhite

13:30

@Dan so it makes sense to send an image hundreds of megabytes large versus pushing a couple of connection settings?

I don't understand that.

Dan

@ewwhite Well, you get the end result exactly as you want it. Not just the settings HPDM deem important

You should spend time configuring the thin client precisely as you want, take the image and then push it out

ewwhite

I did.

I took the image.

I can't push it out because... well, the other thin clients are unresponsive

it took a long time to capture the image

pauska

You now have two people with experience managing HP thin clients, telling you to use images, and yet you go on about it being unneeded

call HP if you're unhappy :)

an (very complex) alternative is to utilize configuration agents on the devices.. afaik they have a SCCM client in the base image, but that will require a lot of work

Dan

@ewwhite Should take about 15 mins all in. Problem is, if you were sitting at the desk you'd see the time it takes for the agent to start doing shit (It pops up a message when it gets a job), reboot, enter the imaging environment and do its business. It's horrible doing it blind in my experience, but you have no idea what the machine is up to

ewwhite

@pauska I took an image, but it's a fair question.

pauska

13:34

I know that the SCCM client can disable the write filter to get settings in.. never tested and never tried

ewwhite

it doesn't make sense if the range of settings changes are a connection string and some local environment settings.

The point is to be able to have a hands-off deployment option for my customer...

they're going to onboard another 60-70 devices and it would be nice for them not to have to manually configure each

Dan

I'm not saying you can't do it with the connection settings jobs, I just wouldn't bother

Whatever is causing your issues is obviously a problem, but I still maintain image management is the way to go. I've imaged 2 - 300 machines over night and walked into a full up and running environment

ewwhite

@Dan I did capture an image. How do I make the initial connection of a new device obtain that image?

e.g. customer should not have to go to HPDM for this

Dan

@ewwhite Tools -> Rules -> First Contact I think

I can't quite remember how that plays, but ensure it does an agent update first

ewwhite

agent update was an easy task

done manually...

Dan

13:42

Yeah, you can do agent update on first contact too

Ideally you want Boot -> Discover HPDM -> Agent Update -> Image Update

ewwhite

@Dan well, I'd like that

Why doesn't this show offline devices?

or is that the reason there's a warning sign?

BigHomie

2 more rep points and I can create tag synonyms

Dan

@ewwhite I think that's what it means

THe little lock in the bottom right indicates the write filter is enabled

ewwhite

and do i want the write filter on or off?

I'm back to this error...

Dan

@ewwhite I'd leave it on by default

ewwhite

13:48

but if I need to make a settings change... it has to be off, no?

pauska

that's when you redeploy a image

Dan

@ewwhite HPDM will deal with it for you

pauska

or use a agent

ewwhite

so now, two of the three devices are up

the middle device has the STOP error... so it may just be bad

I'm rebooting the one with the odd licensing error.

VNC shadowing is working properly

cole

Sup

Dan

13:51

@ewwhite Get the code, but i vote h/w

ewwhite

okay, so the one with the licensing error... I'd like to reimage it.

using the image captured from the working device.

Dan

When taking the image you should have ended up with a template being created

Double click it and point it towards the device you want to reimage

ewwhite

yes.

so it's starting

so i just wait?

pauska

It takes a long while

ewwhite

how long?

Dan

13:55

A least 15 mins or so

Some of the time is just waiting for the agent to pick it up and execute

ewwhite

and there's no other way to view progress, correct?

I also have three HP ThinPro devices... (Linux). Should I even bother with them?

Dan

Not to my knowledge without sitting in front

pauska

t510's use way longer than 15 minutes to flash from a usb stick

ewwhite

they're cheaper, but I don't know if they're necessarily better.

Dan

@pauska But this is using FTP :)

@ewwhite Like I say, they're useless to me ebcause no flash offload

pauska

13:57

@Dan is that faster than a local (fast) USB stick?

Transcript for