Yes, the nodes share images. However, i gave the nodes their own ~/.steam directory, so it's always the same machine that writes the clientregistry.blob file. Sure, i could remove it each time a server is fired up, but it feels like a very ugly hack. :/
The .so files i have in the image directories are: core_i386.so engine_amd.so engine_amd64.so engine_i486.so engine_i686.so filesystem_stdio_amd64.so filesystem_stdio_i386.so libSteamValidateUserIDTickets_amd64.so libSteamValidateUserIDTickets_i386.so proxy_i386.so steam_api_i486.so steamclient_i486.so test1.so test2.so test3.so tier0_s_i486.so vstdlib_s_i486.so I have no idea what those test*.so files are and if they're supposed to be there or not. > Suns grid software. Is there a seperate server image to be run on each > node? or does it share the images and deliver them to whichever node > requires them, as such perhaps running the same image on different nodes > with different ips or hardware ids could be causing the glitch, it is > likely that there is some unique identifier contained in the > clientregistry.blob. > > hows about a dirty hack to get it working temporarily? > add a line to your server initiating script to remove > clientregistry.blob before the server starts ;). > > As to server crashes, bad maps and plugins often cause hlds to crash > unexpectedly and i have not run into any issues when restarting the > server, either scripted or manually, however i have no experience > running the same server image on different boxes. > > Jonathan wrote: >> AFAIK they are all clean halflife installations with only cs running >> (this >> is not hl2 with cs source) >> >> The strange thing is that it all works fine for a few times if i delete >> the clientregistry.blob file, but then after a while the crashes start >> appearing again.. >> >> The way my system works is that a user on my web site "books" a server >> for >> a certain period of time. When that is done, a column is added to an sql >> database. A cron script on the master cluster server reads that sql >> table, >> and if there is a new booking to be made, launches a server on one of >> the >> nodes as a job submitted to the cluster. If the booking time for the >> server has went out, the server is shut down by the master server, ie, >> it >> terminates the submitted job. Does the hlds server need to quit >> gracefully >> in order for it to work properly the next time it's started? I mean, >> does >> hlds recover badly from crashes? >> >> I don't know how sun grid engine works internally.. if it kills hlds >> properly or not (i think it does). >> >> I also have a script that updates everything to the latest versions, >> which >> i run regularly. Still, i want it to work properly the normal way. >> >> >>> Im no expert but it sounds like a bad plugin, what addons do you have >>> on >>> that server? >>> >>> Jonathan wrote: >>> >>>> Right, that got rid of the errors ... for a couple of minutes :( >>>> >>>> I still get crashes, but instead of abort traps i get: >>>> >>>> ./hlds_run: line 303: 7169 Segmentation fault (core dumped) $CMD >>>> >>>> What can be causing this? It still seems to be steam that's crashing, >>>> but >>>> i really can't find out why. Does it have something to do with running >>>> multiple servers on the same box? Can i disable the automatic update >>>> and >>>> instead schedule something as a cron job or such? >>>> >>>> The automatic update also takes some time to complete. >>>> >>>> >>>> >>>>> The problem was that the ~/.steam/ClientRegistry.blob needed to be >>>>> deleted. I tried to delete it for the incorrect user a couple of >>>>> times, >>>>> but when i did it correctly i got rid of the annoying update crashes. >>>>> >>>>> >>>>> >>>>>> Hello, >>>>>> >>>>>> I maintain a network of around 100 counterstrike servers with >>>>>> roughly >>>>>> half >>>>>> of them >>>>>> as bookable servers in a cluster, for a large cs community in >>>>>> Sweden. >>>>>> >>>>>> I run all of the servers on linux gentoo, some on i686 and some on >>>>>> amd64. >>>>>> >>>>>> The cluster software I use is sun grid engine 6, and all the server >>>>>> nodes >>>>>> are diskless clients. >>>>>> >>>>>> The ordinary servers run fine, but the bookable ones crash >>>>>> momentarily >>>>>> with the >>>>>> following error: >>>>>> >>>>>> --------------------------------- >>>>>> AssertPreCondition >>>>>> Name: NonNullFilePath >>>>>> Expr: !sFile.empty() >>>>>> Line: 334 >>>>>> File: ../../../Common/Misc/FileUtil.cpp >>>>>> ./hlds_run: line 303: 8855 Aborted (core dumped) >>>>>> $CMD >>>>>> --------------------------------- >>>>>> >>>>>> I browsed the forums and mailing list and found people with similar >>>>>> problems from >>>>>> late 2005 running FreeBSD 5.4. Their answer to the problem was to >>>>>> update >>>>>> cs using >>>>>> the steam updater. I did that for all 60 cs installations I got on >>>>>> the >>>>>> bookable >>>>>> servers, but they still crash with the error above. I tried to >>>>>> examine >>>>>> the >>>>>> core >>>>>> dumps and strace'ing the processes but I didn't find any useful >>>>>> information. >>>>>> >>>>>> Regards, >>>>>> Jonathan Selander >>>>>> >>>>>> >>>>>> _______________________________________________ >>>>>> To unsubscribe, edit your list preferences, or view the list >>>>>> archives, >>>>>> please visit: >>>>>> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >>>>>> >>>>>> >>>>>> >>>>> _______________________________________________ >>>>> To unsubscribe, edit your list preferences, or view the list >>>>> archives, >>>>> please visit: >>>>> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >>>>> >>>>> >>>>> >>>> >>>> _______________________________________________ >>>> To unsubscribe, edit your list preferences, or view the list archives, >>>> please visit: >>>> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >>>> >>>> >>>> >>> _______________________________________________ >>> To unsubscribe, edit your list preferences, or view the list archives, >>> please visit: >>> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >>> >>> >> >> >> >> _______________________________________________ >> To unsubscribe, edit your list preferences, or view the list archives, >> please visit: >> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >> >> > > > _______________________________________________ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
