Forum OpenACS Q&A: Sysadmin meeting at Valencia
I think it should be a good idea to have an informal meeting between sysadmins at Valencia's OpenACS Conference.
It could be a good oportunity to know us, talk about our experience administrating oacs, our problems, clustering... and maybe talk about future projects (like apache migration, for example), or even start one
It could be fine if at least Innova people (Marcos and Joaquin) and Valencia's (Agustin) could be there, but it can be even better if we can make an international sysadmin group!
I think of it is a very good idea.
See you next week!
I've looked at Apache migration, starting with the old aD code for ACS 3x that ran with Apache 1.3. It could probably be resurrected though the amount of engineering effort required to make it production-quality would be fairly large.
Both of these options are things that might be fun to talk about in Vienna, even informally over beer at night if there's no time available during the days.
I don't worry too much about the lack of a recent unix release if it's rock-solid. If it works, don't break it, no problem. However I don't know of anyone using it for production OpenACS 4.x sites.
The current version of OpenACS does not work due to various reasons, but it is an effort of - i think - less than one week to get it running. Neophytos is running phigita.net on a snapshot of naviserver, but he is not really using OpenACS, but a strongly modified clone. Naviserver is well maintained but a moving target. I said a year ago or so that i will try get OpenACS running on Naviserver, once there is a new release as a reference point.
I do not see a big problem with the current situation of Aolserver. If a real problem shows up, or maintenance becomes a problem, we will move our apps to naviserver.
Going to apache is a long way, and i am not sure about the advantages, esp. when c-code modules from aolserver are involved.
i am not sure about the advantages, esp. when c-code modules from aolserver are involvedIt just seems to eliminate a lot of other moving parts; and it would make it easier for OpenACS to fit in a standard LAMP box, since aolserver seems to be the hardest thing to keep running...
19th afternoon seems to be free, after the .LRN consortium presentation ( http://aulavirtual.uv.es/ficheros/view/agendax18x19.htm ).
We can have the meeting then, and go dinner after that
Wouldn't a meeting targeting and inviting "sysadmins" be an appropriate place to compile and share "configuration profiles" for OpenACS/AOLServer deployments rather than discussing the highly improbable chance to have it ported to what-ever-platform-technology?
I mean, there are recurring threads on these forums on the voodoo of aolserver configuration:
- How to appropriately trace, identify, and cope with performance issues? Why not compile all the useful infos repeatedly given by Gustaf N., Dave et al. nicely and give some tool support (extend developer-support, xotcl-request-monitor)?
- How to DOCUMENT and REPORT them so others can help effectively?
- Provide two or three sample configs, as available for other platform technologies (RDBMSs, ...) that target different (organisational) users. Or, a web-based configurator where someone can enter profile attributes in a guided manner?
- Provide a repository to store and share configurations?
This is even something you can get started at the conference: in medias res ...
As for Apache: Do not underestimate the implications. Apache has different concurrency models and the infrastructure is sufficiently distinct. Btw. if you look for stable or responsive maintainership, mod_tcl is not the best choice ... though there were life signs at this years Tcl2008: http://www.tcl.tk/community/tcl2008/schedule.html (Olly Stephens talk)
Enjoy your time in valencia!
Instead, (and admittedly in my humble opinion), I would recommend looking at a Java model.
There is already a port of TCL to Java, built in JDBC connectivity, interfaces to any web server out there. Java has an internal model much closer to AOLserver in terms of threading, shared memory, etc.
The main purpose of the sysadmin meeting is to meet each others and share some knowledge that is not in the forums right now, talk about our common problems and solutions for them, all in an informal way.
A lot of this common problems are related to the use of aolserver. and I think that the posibility of an apache|naviserver future migration is an interesting topic to discuss. Aolserver seems to be low maintained, specially some of the modules (last commit on nsopenssl was on 2006, and the last one on nsldap and nscache is 5 years old). Openacs is fully dependant on aolserver, and this can take us into problems in the future.
About the topics you are pointing out, it would be great that we could talk about all of them too. They seem really useful to me, and can be achieved in a short period of time :)
It's a pity that you will not be in Valencia to talk about all this things while having a beer
>> Are you using NaviServer in production
>> environments at Qwest?
We're not, not yet! and we'd be also interested on Apache
We've been doing some research about moving to Apache and we've contacted the guys working on portable-nsd project but we didn't get any feedback. Had you seen that project before? They seemed to have a plan to run oacs under Apache and looks like thay've done some work on this.
Getting a robust implementation working under Apache would be a large implementation effort. My background is in compiler writing and operating systems work (mostly old ones, like the original 1974 version of Unix from Bell Labs), so I don't see it as impossible, but certainly not trivial.
Don't worry, I won't sugar coat the effort involved.
As far as what sysadmins should talk about at an informal meeting held at the conference, I'd suggest they should talk about anything they'd like to talk about.
Yes, I've looked at the portable-nsd project. Not really the way to go, IMO, for production quality sites.
How to appropriately trace, identify, and cope with performance issues? Why not compile all the useful infos repeatedly given by Gustaf N., Dave et al. nicely and give some tool support (extend developer-support, xotcl-request-monitor)?
Going back to Stefan's list and the nearer term. Would someone be able to do a sysadmin session with a dicussion focus on the first two items on the list?
A meeting is a wonderful idea (and I'm planning on flying in even if I'm only a reluctant pseudo sysadmin at the moment) but a more formalized almost-tutorial session would be a great addition...
Though I've only been a pseudo-sysadmin, I like aolserver. It's stable and mature and doesn't keep changing like Apache. We can build applications on top of it, rather than being concerned with the internals of the web server.
If aolserver needs a change that we require, can't we make it, and commit it? It's hardly "not maintained" if it's open source.
What performance issues exist out there for the big implementers?
Finally, didn't ARSDigita go bust when it tried to move ACS from Tcl/AOLServer to another platform?
Just my two cents.
If I'm forced to move from Tcl, I'll be a lot more interested in moving to a language like python than Java.
The problem is misguided, poorly thought out, undocumented changes to the code that in the past have been undertaken without public discussion, or even public notification.
And an attitude on the part of the guy who thinks he "owns" AOLserver, and there he can do whatever he wants without regard for the community.
I still use 4.0 not 4.5.
Daveb should upgrade the site for us :)
Well, if my misplaced responses confuse you, ask me about them in Valencia!
See you there!