> -----Message d'origine----- > De : Gaetano Giunta [mailto:[email protected]] De la part de Gaetano Giunta > Envoyé : jeudi 26 mars 2009 17:35 > À : Sébastien MOREL > Cc : [email protected] > Objet : Re: [Sdk-public] Cluster in eZ 4.1 - Important points > > Sébastien MOREL a écrit : > > -----Message d'origine----- > > De : Gaetano Giunta [mailto:[email protected]] De la part de Gaetano > > Giunta > > Envoyé : jeudi 26 mars 2009 16:08 > > À : Sebastien MOREL > > Cc : [email protected] > > Objet : Re: [Sdk-public] Cluster in eZ 4.1 - Important points > > > > Sebastien MOREL a écrit : > > > >>> Hi, > >>> > >>> I just upgraded my eZ 4.1 in cluster mode. And I found several > >>> important points for the next version: > >>> > >>> - Clusterize.php needs mysql_ functions, so clustering is impossible > >>> while using only mysqli. > >>> > >>> > >> excuse, but it seems to work fine when oracle is in use (even though I > >> might not have disabled completely the php mysql extension). Can you > >> point me to the dependency? > >> > > > > The fatal errors came from : > > kernel\classes\clusterfilehandlers\dbbackends\mysql.php > > > > > Correct. > In fact there is no eZP cluster driver yet for the mysqli db driver. > >>> - The FileHandler definition have changed. Only the settings/file.ini > >>> describe it, dont forget to add it when the 4.1 doc will be ok. > >>> > >>> > >> I think that is described in the BC docs. Of course it should be in the
> >> upgrade manual, when that is online > >> > >>> - The DBBackend has changed too. But, the index_image.php still uses > >>> mysql , its a bit confusing. > >>> > >>> > >> Good catch > >> > Opened an issue for you: > > http://issues.ez.no/IssueView.php?Id=14674&ProjectId=3 > > Otoh I think that changing the name of the constant in index_image.php and thus renaming the existing file not really worth the hassle in the end... I agree with you, I think, that the include of index_image_DBBACKEND.php into the index_cluster.php needs to be upgraded. A poor solution, but a coherent one, would be to have a new file named : index_image_eZDBFileHandlerMysqlBackend.php The choice between mysql or mysqli should be implicit with ezdb. > > > >>> Note that, if you define the FileHandler with mysql in the file.ini > >>> and you run the clusterize script, the whole instance is broken. I > >>> havent found the solution to fix, expect by using a backup ;) > >>> > >>> > >> Care to explain better what results you get? SO that we can try to make > >> the process less destructive? > >> > > > > The clusterize script stops at this error : > > "Clustering settings specified incorrectly or the chosen file > > handler is ezfs" > > > > I found out that > > $handler = eZExtension::getHandlerClass( $options ); > > Return false in : kernel\classes\ezclusterfilehandler.php line 65 > > (instance() method) > > > > I tested this 3 times with a fresh backup of my ez instance. > Which is correct. As there is no cluster driver for mysqli the script > should stop there with an error. > > Each time after > > the error execution, the whole instance was broken. > > > This is otoh totally unexpected. > > To fix my instance I cleaned the cache folder manually but nothing. > > > > The errors (when my instance was broken) was the default exception error: > > An unexpected error has occurred. Please contact the webmaster.. > > > > Are you sure you are not having a problem in the main db connection > configuration? > The error message in fact seems to point to this: now an exception that > is thrown when connection to the db cannot be established (the main db > connection, not the one used by the cluster handler). I don't know exactly, I'm sure at 95% that I deleted each time the file.ini and emptied manually the cache to fix the problem. But it didn't fixed anything. I remember I had an error.log message about ezdb class (I think), but unfortunately I deleted the log file... I didn't import the database each time (so the error didnt come from a modification of database) Here are the folders emptied manually : Var/cache Var/PROJECTNAME/cache What is clear is that : The instance worked ( I cleared the cache, N times with purge or not and it still worked) I configured the file.ini (with an error in the FileHandler) and I executed the clusterize.php The clusterise script stopped and the instance was broken When, I have more time, I will retry to give you more details. Perhaps, I forgot some steps..., perhaps the file.ini wasn't deleted at the first test of my instance (load of the home for example) and thus the problem doesn't come from the clusterize.php but come from a loading of a page with a bad file.ini... A fix can be to stop the execution when a FileHandler doesn't not exist. Thanks for your answers > > I didn't investigate much because I was short. > > > > > >>> Otherwise, when you know this points everything is OK and works > >>> perfectly ;) > >>> > >>> I hope these notes will be helpfull. > >>> > >>> Regards. > >>> > >>> > > Regards > > > > > > > > -- Sdk-public mailing list [email protected] http://lists.ez.no/mailman/listinfo/sdk-public
