Teal Shoes Are The New BFG 9000

[serinde]

I am on the world's most dire conference call, which is for all CUNY Help Desk managers (and the woman at the Mothership who fills that role). We are a half-hour in and so far it's been all about the incoming Oracle PeopleSoft bleargh. Not a word about how students couldn't register for six hours on Monday. But I don't care, because I am wearing new clothes and have borrowed nedlnthred's awesome teal shoes. I feel like I can throw down with anything.

OH WE JUST GOT TO THE PORTAL ISSUE. "It's not down! People could still get to Blackboard! It was only LDAP!" (WHICH YOU NEED TO REGISTER WITH.) Oy veh es mir. "We see usage spikes and that is what causes the problem, we don't know what's causing that and are looking into it." Gosh, d'you think that maybe, possibly, they could be coincident with registration sessions?

Comments

[blarglefiend.livejournal.com]

My experience on the other end of these things (i.e., as one of the guys responsible for it continuing to work) was that there was always a tension between management's desire to not spend outrageous amounts of money -- yes, this while buying Oracle products -- and the unfortunate reality that you get these peaks maybe twice a year that are so far beyond the "normal" load.

Of course with LDAP you also get the bonus that really wants to be a read-only database. Writes are horribly inefficient. So I can easily see that keeling over under the load of registration day if the service simply hadn't been scaled to handle that sort of load...

(Funny how we never had these problems when we were using the custom in-house system running on a single Alpha box. Hah.)

[spride.livejournal.com]

There are no writes to CUNY LDAP during registration. All it's doing is authentication to the Portal, whence they access SSO apps such as the registration app. Spikes in LDAP auth transactions have been a fixture of registration season since Noah was a kid. No excuse.

[blarglefiend.livejournal.com]

Fair enough: this was just an educated guess from having worked in similar environments.

I still wouldn't be surprised if the underlying problem were that the LDAP infrastructure had been scaled to deal with the non-peak periods and simply doesn't have the capacity to cope gracefully. That's the sort of bonehead move I'd expect.

[spride.livejournal.com]

From what the CUNY CIO said at an unrelated meeting yesterday, they lack front end servers to handle the load. They're using Vignette.