Re: The way Apache is restarted and the WO adaptor
Will Scheidegger <[email protected]>
| Newsgroups | gmane.comp.web.webobjects.admin |
|---|---|
| Message-ID | <[email protected]> |
Stefan, Laura Our setup on a pretty old G4 has been running pretty stable for quite some while. And it's not easy to say exactly when things started to get ugly. We continuously installed more apps on the server, but the the processor is still more than 75% idle. Fact is that our apps become unavailable quite often now although they're still up and running. Sometimes the wo adaptor acts up (we think this started when we activated mod_rewrite on one of our sites) and more recently wotaskd seems to go crazy. But... - we could not track down to problem yet - we did not find a proper way to restart WO without restarting the whole box - every time we think we might have found a symptom (like too many CLOSE_WAIT connections listed by the lsof command) it turns out that we cannot use it to reliably watch the server... - in short: We have no clue how to resolve these problems. We decided quite a while ago that we wanted "real" server hardware and that we would go with Solaris (our other servers are already SUN boxes). So it's not just the recent problems with WO deployment that drove us to that decision. As I said before: I really would like to have a couple of Xserve four our deployment, but it's not really what we need and when it comes to deployment I have to respect other opinions to in our company ;-) -Will On Friday, January 9, 2004, at 06:06 PM, Stefan Apelt wrote: > Hi Will, > > your mail from 09.01.2004, 17:47: > >> Damn! We just bought a Sun server because we gave up on WebObjects on >> MacOS X Server because of problems like the one you're describing. Now >> you're telling us that wotaskd / wo adaptor is flunky on Solaris too? >> Very encouraging. I think I'll move to Tomcat! > > I don't know about the differences in JVM implementation for Solaris > vs. MacOS vs. Linux. We have Linux 2.4.x+Java 1.4.x+WO 5.2.x running > for around a year now and wotaskd never made problems (1 web server, 2 > app servers, 1 db server, ~20 apps, ~50 instances, lots of requests, > servers are connected via switched Gigabit ethernet). I did not try > JDK 1.3.1 on Linux but with 1.4, it works very well. I don't know > about Linux on Sun either, for that matter. :) > > By the way, we almost never restart Apache nor wotaskd (well, three > times a year or so for upgrading). Is this really necessary? Or could > it be a performance or timing issue, eg. that you have a slow network > connection between servers, broken cable, poorly written network card > driver or something like that which drives wotaskd crazy? In that case > a new server might not help much... > > -- > Regards, > Stefan >