Thanks for replying Aaron. I was passing the correct information to the adaptor. It's done via environment variables in the script that launches the adaptor. I am not certain what the problem was but the combination of deleting the socket and also deleting a file in the /tmp directory that the adaptor was creating and never deleting. Even when receiving a TERM signal the cgi-adaptor doesn't seem to clean up the file in the /tmp directory. I'm not sure if that is intentional or not but everything seems to work as expected now. I'll dig more into it later.
The other issue of not being able to shutdown apps in a multiserver setup was solved using a property to let Wonder know the LAN ip addresses to pay attention to. It's recommended in the wiki somewhere for VPS setups that have multiple IP addresses. er.extensions.WOHostUtilities.localhostips=(ip address) And finally, I put womonitor and wotaskd together on one of the machines. It was not clear to me at first that womonitor had to run on a machine with wotaskd. I assumed wrongfully that since wotaskd could run standalone that womonitor could do the same but it looks like it needs a wotaskd instance to persist it's configuration. So for the moment everything works and I have a full setup with Jenkins CI running. On 14 Feb 2014, at 00:19, Aaron Rosenzweig <aa...@chatnbike.com> wrote: > Hi Kevin, > > It’s fine that 1085 ports are open… but that doesn’t mean that the fastcgi > adaptor knows how to find them. It won’t multicast across all systems on the > same LAN to find open 1085 ports. > > You need to specify in the fastcgi configuration file where your wotaskd > daemons are running. Something like: > > WebObjectsConfig http://www.server.com:1085,http://www2.server.com:1085 10 > > Sorry, I don’t remember where the configuration file is. Probably have to > look at the source code for it. > > Yeah you’re right to use nginx, it’s the best and uses little memory. > > Here’s a reference document on the “theory” in case you haven’t seen it: > > http://wiki.wocommunity.org/pages/viewpage.action?pageId=7143530 > > Cheers, > AARON ROSENZWEIG / Chat 'n Bike > e: aa...@chatnbike.com t: (301) 956-2319 > > > On Feb 5, 2014, at 9:43 AM, Kevin Hinkson <k...@coralstone.co> wrote: > >> Hi everyone, >> >> I am having some difficulty getting a multi-server setup to work. I was able >> to setup nginx + fastcgi adaptor + webobjects to run all on one machine. A >> multi-machine setup is showing me that I probably don't know enough about >> how the webobjects pieces all work together to set this up. The full setup I >> am going for uses jenkins to build and deploy apps to 2 machines with a >> separate web server running nginx. >> >> Server setup: CentOS 6.4, JDK 1.7, the Wonder JavaMonitor and wotaskd >> >> Machine 1: jenkins, nginx, webobjects fastcgi adaptor, javamonitor, postgres >> Jenkins works fine. The nginx -> adaptor config is setup as directed in the >> nginx config provided in the FastCGI adaptor source and points to a local >> socket. >> >> Machine 2 + 3: wotaskd and webobjects applications >> I can initially setup wotaskd and the applications (via womonitor on machine >> 1) which launch correctly. I am unable to shut them down and each app >> usually, spits out a broken socket error on launch >> >> WebObjects ports opened on the local network: 1085 on the app servers, as >> well as the application server ports (2001, 2002 etc) >> Hostnames: all fqdn setup in /etc/hosts on each machine and used to >> configure javamonitor and the fast cgi adaptor >> >> Problems: >> 1. Since womonitor is standalone (without wotaskd) on machine one, a reboot >> seems to mean it looses the configuration settings, password etc and it >> doesn't seem to read this information from disk on launch since there is no >> wotaskd running on that machine. Even after relaunch and when rea-dding the >> hosts I get a password error. I have to wipe the SiteConfig.xml on the app >> machines to get them to play nice again. Is there a way to avoid any of this >> as I would like to use javamonitor so that deployment can be automated >> through jenkins. >> >> 2. I can hit the AdaptorInfo page but it seems to show stale information >> from the single machine setup. I launch the adaptor using spawn-fcgi and use >> environment variables to pass the config url and user name and password >> which I think are working because I can access the AdaptorInfo page using >> the same username and password and the WOConfig urls on machine 2 + 3's port >> 1085 work (prior to rebooting). I have no idea what else to do here. It's my >> understanding that nginx talks to the adaptor through the socket provided >> and the configuration urls tell the adaptor where to look for the running >> instances and configure itself. >> >> 3. The webobjects applications aren't reachable via nginx. The launched apps >> have no problem connecting to the postgres database on machine one. >> >> 4. Maybe this should be the first problem, but is the FastCGI adaptor >> supposed to work in this kind of setup? I am assuming most people don't use >> it so there is little information I could find on how it should be setup. I >> would prefer to stick with nginx. >> >> Does anyone have any pointers? >> _______________________________________________ >> Do not post admin requests to the list. They will be ignored. >> Webobjects-dev mailing list (Webobjects-dev@lists.apple.com) >> Help/Unsubscribe/Update your Subscription: >> https://lists.apple.com/mailman/options/webobjects-dev/aaron%40chatnbike.com >> >> This email sent to aa...@chatnbike.com >
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Do not post admin requests to the list. They will be ignored. Webobjects-dev mailing list (Webobjects-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
