<$BlogRSDUrl$>

Wednesday, September 22, 2004

FSF problems - Name collisions

I was having a problem with FSF on NDSUSR3. The Engine screen showed "E111 Unable to confirm existence of Directory".

Below is an exchange between Matt and me about how to do certain FSF things:

Session Start (ICQ - 20023056:Mattd): Wed Sep 22 14:18:27 2004
puppy: I got question about FSF when you get a chance.

Mattd: ok. shoot.

puppy: E111 Unable to confirm existence of Directory. A couple of months before you left you went through one of these, but I've forgotten how to get rid of it. Is it referenced in the manual (for some reason I didn't think to look there before bugging you :)*** Auto-response from
Mattd: I am currently away from the computer.

Mattd: have you turned on debugging for that event?

puppy: I'm gonna say no.....

Mattd: ok. go to "Manage FSF Engine" and go to your pending event list. this is in the web gui

puppy: OK, just a minute let me get therepuppy: OK, at pending events.

Mattd: first thing is to look at and examine the name collisions.

Mattd: you can probably delete those, but don't do it yet.

puppy: Under pending events?

Mattd: click on the event for 1_3.student.usr.uky

Mattd: yep

puppy: k

Mattd: click on enable debugging for this transactionpuppy: done

Mattd: then click, view log for transaction

puppy: k

puppy: see a 601

puppy: So the esspea2 user object is the problem.......

Mattd: yeah, that's not the important thing. look at the object the rename is for.

Mattd: yes.

Mattd: make sure that that account exists and that they have a home directory attribute and that they're cataloged.

Mattd: this is easy to do in fsf!!!

Mattd: go back in your browser

puppy: OK, show me how daddy

Mattd: click on the help button

puppy: k

Mattd: click on object diagnostic and enter esspea2.student.usr.uky and click submit

puppy: Has a home path

Mattd: right. so, you should be able to delete this event. because the one that it was originally for is all taken care of.

puppy: OK.....

Mattd: so, go back to your pending events and delete that sucker

Mattd: you'll want to do the same for all the name collision events....those with something_something

puppy: by deleting you mean aborting?

puppy: That was going to be my next question

Mattd: yeah

Mattd: there are a few others that need to be looked at too

puppy: OK, here's one that got triggered, but didn't get finished.....How do I fix one like that?

Mattd: which one is that?

puppy: 0_4

Mattd: notice that you actually have 3 events for clora2

Mattd: that 0_4 is one of them.

Mattd: that's the rename for name collision.

puppy: Hmm,

Mattd: i would delete the rename first.

puppy: OK. But it said something about the create policy....

Mattd: what you're gonna do is go back and individually backfill that user so that they get fixed.

puppy: oh, OK

puppy: So, would I delete all of the records for that user?

Mattd: yeah. i would

puppy: OK

puppy: OK that's done. Now do the backfill for clora2, right?

Mattd: right

Mattd: click "manage fsf policies"

puppy: Then drill down to FSF_Locker and manage it?

Mattd: no. browse to student.usr

Mattd: it'll take awhile for it to build the list, but be patient. that's a lot of users :)

puppy: OK!

Mattd: :)

Mattd: you could do a backfill on the entire container and you might want to do that just see how things are looking.

Mattd: but, you don't have to do that just yet

Mattd: we have a new thing in the backfill where you can run a consistency check. that's pretty cool and may come in handy.

puppy: Well, I was thinking about doing that, but I just had to make a bunch of changes to some users accounts - because John and I never got together to see how we wanted to do the class groups.

puppy: OOOH, that would be handy

Mattd: when you do that list of all the users, it only shows them in groups of 100. so, you'll have to locate the group of 100 that has clora2 in it. they're listed alphabetically.

puppy: No search function ;)

Mattd: actually...a good place to have that would be in the object diagnostic. i'll put that on the list.

puppy: would that make me famous?

Mattd: unfortunately...no.

puppy: "This function dedicated to PG"

Mattd: that's in the easter egg :)

puppy: LOL

Mattd: do you have your list yet?

puppy: Still retrieving

Mattd: k

puppy: got, now trying to find to right one

puppy: Ummm, that user isn't there

puppy: goes from cloqui2 to clord

Mattd: user's not in the tree?

puppy: clora2 should be between them, right......

Mattd: you'd think...

puppy: hmmm

Mattd: do they show up in consoleone?

puppy: hold on, lemme see

puppy: Nope, not there

Mattd: well, that explains that :)

puppy: yep

Mattd: they were probably deleted by JW and the account was never resynced

puppy: ah. What would I have done to backfill just that account?

Mattd: well, once you found it, you'd click on the backfill icon and select the backfill option that you wanted and off you go.

puppy: Which icon is the Backfill one :)

puppy: My little ballons aren't coming up

puppy: Ah, that's a backhole shovel isn't it :)

Mattd: yeah

puppy: so, if I wanted to do the whole container, I would simply do that on the container a few pages back ?

Mattd: yeah

puppy: OK. Well I'll take care of the others. Thank you for the help!

Mattd: browse there and I'll show you some things.

puppy: oh, OK

puppy: OK, I'm there

Mattd: ok. typically for a user backfill, you want to do the top-most backfill option with enforce policy paths checked.

Mattd: when you want to run a backfill on the container, i would do that one.

puppy: OK

puppy: But I should do it in Test mode first, right?

puppy: Apply mode that is

Mattd: yeah. just to make sure everything looks good and its not like moving all your data to a hospital server or something like that :)

Mattd: check mode

puppy: LOL, OK. Now the Perform Consistency Check will just give me a report of how things look?

Mattd: yep. that does a check to see what ds and the file system thinks compared to our dbase.

Mattd: pretty cool

Mattd: the format in HTML needs some work tho.

Mattd: I guess that's my job :)

puppy: So I can do that periodically without any problems

puppy: I guess so, get on that!

Mattd: surepuppy: Hmm, neat.

puppy: Anything else I should see while I'm here and paying you?

Mattd: in your pending event list you have some events that say undefined action. i'd look at those and see what's up.

puppy: OK. I did want to go back and look at all of those.

Mattd: we have a new version of the engine that allows you to backfill groups.

puppy: That will be nice when I get my groups going

Mattd: yeah. i've got the code, just haven't done a build recently.

Mattd: there's also some bug fixes in there for groups. oh!!!

puppy: Esplain the event queue on the server. Is that how many are still pending......

Mattd: you can also set quota's on directories inside group directories now.

Mattd: that's pretty cool.

puppy: No way!!

puppy: That's awesome

puppy: That freaking rocks

Mattd: yeah...so if you have say...pglenn inside a group home directory, you can set the quota on pglenn

puppy: That would be soooo cool

Mattd: pending events tells you how many events are currently pending.

Mattd: having pending events is OK. especially when they're pending deletes.

Mattd: you periodically want to check your event list/queue to see if there's anything that's not processing properly.

puppy: Oh, OK, I was wondering why they weren't going down. But that esplains it

puppy: OK.

Mattd: yes. those won't go away till their 90 days are up.
Mattd: so, they just hang around till its time for them to be deleted.

puppy: So, with the groups.....you can set a quote for the entire directory, then the subdirs ??

Mattd: if you so desire

puppy: Oh, now that's so cool. I'd have to do that or one bozo will fill up the whole thing

Mattd: yep

Mattd: those quotas are set on the template directories for the group objects.

puppy: That concistency check is pretty cool too. I can only assume the groups would be listed to, right?

Mattd: yes

puppy: coo

Mattd: if they were managed

Mattd: there's a way to delete multiple events from the event queue, but i don't think you'll need that.

puppy: There are bew just a few.

Mattd: we don't publish that. we don't want people to do stuff like delete all their events then complain to us that FSF isn't doing its job :)

puppy: LOL, they wouldn't do that.......

Mattd: oh...they would

Mattd: we've had some pretty weird things done.

puppy: I can only imagine

Mattd: like, using dirxml, accidently renamed every user to the wrong format.

Mattd: all sorts of stuff.

puppy: Whoops!

Mattd: yeah.

Mattd: not pretty

puppy: Oh, crap.

Mattd: ??

puppy: No, I was just talking about what you were saying

Mattd: :)

puppy: You said anything with something_something should be looked at right?

Mattd: anyway, i think that's about all their is...

Mattd: you know about the quotamanager?

Mattd: what about the trustfun stuff built into the gui?

puppy: For the directories?

puppy: in the policy?

Mattd: yeah. where you can have someone else manage quota for users.

puppy: Trustfun is in this too? That's cool

Mattd: i'm not sure you would need that.

Mattd: yeah.

puppy: No, didn't know that you could designate that too

Mattd: yeah. that's pretty neat feature

puppy: I'll take a look around and see what else I can find. Then I'll prolly have more questions

Mattd: ok.

puppy: I'll let you go for now, thanks

puppy: !!

Mattd: a good thing to do is to browse to your user and click on the folder icon for it.

puppy: Tell Dorothy to give you a kiss from me!

Mattd: it shows you how much data you've got in your dir. the types of files.

puppy: OK, I''l do that right now

Mattd: it also shows the rights someone has to the dir and how they get them.

Mattd: (k)

puppy: Right back atcha

Mattd: check out the red N icon. click on it. that's a pretty cool thing.

puppy: ok

Mattd: we need to find a way to speed up the list of the users.

Mattd: not sure what we can do tho. we're kinda bottle-necked by DS and the file system.

puppy: don't know how to speed up 70+k

puppy: Yep

Mattd: (Link: https://ndsusr3.uky.edu:8009/FSF/HTTP_Environment?6F54F1695D163332C389B9A446D1A92DPGLENN.STUDENT.USR.UKY)https://ndsusr3.uky.edu:8009/FSF/HTTP_Environment?6F54F1695D163332C389B9A446D1A92DPGLENN.STUDENT.USR.UKY

puppy: Hey, now that red N is cool

Mattd: yeah

Mattd: anyway, that's just a few things to play around with.

puppy: Well, thanks again.

Mattd: np. lemme know if you have questions about those pending events.

puppy: OK, will do

Session Close (Mattd): Wed Sep 22 15:59:46 2004

Session Start (ICQ - 20023056:Mattd): Wed Sep 22 15:59:47 2004

Mattd: some of them look like the user was deleted and then recreated.

puppy: Yeah, I think that's what is up with that first one even

Mattd: in those cases, i guess you could delete the pending events for both of them, then backfill the user.

puppy: But I'll have to look at things closer

puppy: OK, that sounds like a plan. I'll give it a try.

Mattd: if the account was recreated, you'll definitely want to delete the pending delete. else it'll delete their stuff :)

Mattd: anyway...good luck.

puppy: OK. Thanks again Matt

Wednesday, September 08, 2004

Last Friday the web interface to Locker went down. The service is provided by Student2. Apache was loading on it without any errors. The Cert was renewed in August and the adminserv.conf file was updated. The problem turned out to be a line in the Autoexec.ncf file. After updating this things started working out just fine.

Last Friday evening the web interface to Locker went down. Apache was loading without any errors. The Thawte cert was updated in August. Matt called me back and finally figured out it was a line in the Autoexec.ncf file on Locker that needed to be updated. I had updated the adminserv.conf file on Student2, however had not updated autoexec.ncf.

This page is powered by Blogger. Isn't yours?