Archive for September 2003
I completed moving all (relevant) sources to CVS. Today sla management, snmpPoller (perl), netcoolAdapter (perl) and rrdtool (C) were finalized.
I agree with your analysis.
With Objectserver, we already have the “store” portion of this in place. All that is needed is a simple netcool automation that checks for events that sit too long in the “raised” state, meaning that an ack never came back from the tools, and then reraise those events.
Granted, this is application level store and forward and not middleware level, but regardless, it means, as you correctly point out, that the gateway code is simplified.
Events often come in bursts. Bursted events must flow unimpeded between Dispatcher and Micromuse. In order to support this, buffering or queuing of events is needed in the receiver. The current single-threaded HTTP stack in the Perl eventAdapter cannot do any buffering.
The problem with cryptographic errors during the build due to a
corrupted keypair.snk is solved. It turned out to be a CVS problem. The
.snk extension was not marked as binary, so that CVS corrupted the file
(probably due to macro expansion). It was removed and readded after the
cvswrappers file was adjusted. Note that simply removing is not enough.
It needs to be physically removed from the CVS Attic on the source
We finally got the nAnt-based build process to go cleanly today. The
goal is to get and keep the build working in a virgin CVS sandbox.
In order to make this as painless as possible, the autogenerated web
references in Visual Studio.net are added to CVS as well. In principle
it is not a good idea to directly add generated code. Instead the
generator should be invoked during the build. However, the generated
web references are pretty brain-dead (they include a hard-coded URL) so
that in almost all cases they need to be modified anyway.
The better solution is probably to subclass them and check-in the
subclass only, but there is still a chance then, that the build will
brake. For now, we feel we can live with it.
Another problem is external dependencies. Right now, they mainly, and
most problematically exist in some of the tests. The SSL-VPN code tests
all rely on a working LDAP-SSL connection to the AD server. This might be a problem in TXS, although when using
LDAP over SSL, windows is not supposed to make sneaky COM calls
anymore, but stick to port 636 for everything. Another problem is the
dependency of the Syslog Viewer tests on the presence of syslog
files in a certain format outside the source tree. This can, and should
be resolved easily.
For the time being I am still using my existing worklog. The RSS feed is at [censored]
The weblog itself is at [censored]