More strange issues after upgrading from bind8 to 9

[Kevin Darcy]

Are the machines simply getting overloaded? If you don't take advantage of the
multi-threading capabilities of BIND 9, I imagine it could actually perform a
little worse than BIND 8 in some situations. What do things like "sar",
"iostat", "vmstat" and so on tell you about how the machines are performing?


- Kevin
Richard Whelan wrote:

> Dear All,
>
> A couple of weeks back I mailed regarding crashes we were getting after
> upgrading from 8.23 to 9.1.3. We are running on Solaris 8.
>
> This was happening on all three of our systems. Last week I ended up taking
> one
> of the systems back down to 8.2.4, after seeing yet another strange problem.
> The system that crashes, is currently with ISC, but the other two were
> showing major timeouts.
>
> I know there has been loads of messages regarding the timeout messages,
> which basically indicate either a zone or a network issue. However, what we
> are seeing doesn't seem to fall into either of these. The two systems, which
> are
> a master and slave, are located in two different buildings. We have
> our own network, and have been ensured that no work has been carried out at
> the point when the problems started to occur. There are no firewalls, or
> anything we can think of that may be in the way causing an occurrence which
> is turning out to be almost regular as clockwork.
>
> On the slave we start to see the refresh/timeout messages, but they are the
> address of the other master and other masters which we are also slave to.
> Initially they are not causing much effect, but after a little while, the
> intensity of the messages becomes almost constant. The message log on our
> master seems to stop logging almost all messages, and appears to go to
> sleep. Now customers are now noticing timeouts in resolving, and end up
> calling us. At this point we have to do a restart on the process, on both
> systems. They both appear to be caught up in what ever is happening here.
>
> Since down grading the master back to 8.2.4, on Thursday last week, this
> machine is no longer crashing, or showing the timeouts mentioned above. The
> Slave, however, although not timing out to our Master any more, is still
> showing the same symptoms.
>
> If this happened only the once, I could possibly understand it, but as I
> said, it's as regular as clockwork, as it happens again almost 4 1/2 hrs
> later and so on. Our NOC are having to kill the process still.
>
> I know alot of people have seen the same error messages, but has anyone
> noticed a degradation in service at the same time, or any regular patterns
> to these messages being logged ?
>
>