Leslie Clark wrote:
>
> In my experience, the 5.1.1 patch cleared up a bunch of problems
> with incorrect status in Collections. There have been a couple of times
> lately where I thought I saw something with incorrect status in a collection,
> but was unable to recreate it. Usually it happened if the box was under
> stress, so I chalked it up to nvcold or nvcollmap being behind, and
> being impatient, used Ack/Unack to get it to pick up the current status.
> But I am only at the customer site for a week or two, and may not see
> things that crop up later.
>
> Are you sure that the thing in the collection and the thing in the IP map
> are the same object? (ie they have the same object id, you select one
> and both get selected?
I'm almost 100% certain that there were single instances in the
database, but the first time was 3 weeks ago. I'm usually do a bit of
probing in the databases with ovobjprint and ovtopodump, any time I
notice something funky like this.
>
> Does it look like the node placed in the Collection came from one of
> the network submaps rather than from the IP Internet submap? I mean,
> you can get a red router with red and green interface cards under it,
> and status source is symbol rather than a yellow router with status source
> of compound? I think this used to happen a long time ago but I have not
> seen it lately.
I had hosts, hubs, routers, and even networks (yes, I had a collection
of related subnets. isNetwork=TRUE and some regexp for the IP
addresses). All of these different types of objects exhibited the
behavior. I can understand how the hub or routers could be confused,
but not the host (which was singly homed, non-forwarding) or the network
symbol (there's only one place to get the status for this).
Another oddity though was the fact that often the incorrect status was
blue or brown.
> There also used to be some confusion between the node and its interface,
> depending on whether it was discovered by address or by name. Deleting
> the node and discovering it by name cleared that up, I think.
>
Well the first customer had real thin DNS coverage and almost none on
their routers. As before, I can buy this argument with the routers,
but not the other objects.
I'll track it down don't worry. I'm looking around this customer's
setup now. And will keep some of this in the back of my mind.
chris.cowan.vcf
Description: Card for Chris Cowan
|