Multiple servers without shared disk? Will this work?
Permalink
Please let me know if this theory holds water. Thanks in advance!
As it stands today, our "cluster" of C5 nodes uses NFS to host all of the various websites, all out of one NFS mount. For various reasons, we're being asked to remove NFS from the equation.
I *believe* we can use local disks on each node to hold the web info, and simply use rsync to keep the file structures synchronized from one "master" directory onto each local set of disks.
It does raise the issue as to whether /my_website/files/tmp/ directory should be replicated to individual machines or not. Just to make it more interesting the /my_website/files/cache directory is also changing randomly - as in directories and such coming and going.... If someone has helpful hints for actions to take post-rsync, please let me know For example: is it ok to stop services, delete everything in the /cache/ directory and restart? How about /files/tmp/ directory?
The mysql service will continue to be hosted on the network, so that's a non-issue.
After reading through how people handle "centralization", it looks like the /concrete in each web directory is meant to simply be kept at the same version, nothing more. There are no edits going on in that directory (one hopes!)
So, if all of my DocumentRoots are rsync copies of the master volume, and we carefully plan any C5 software updates, and edits to C5 sites are only done on a Master instance, then we should be in business, yes?
Suggestions/caveats/speedbumps/glaring errors are all welcome.
As it stands today, our "cluster" of C5 nodes uses NFS to host all of the various websites, all out of one NFS mount. For various reasons, we're being asked to remove NFS from the equation.
I *believe* we can use local disks on each node to hold the web info, and simply use rsync to keep the file structures synchronized from one "master" directory onto each local set of disks.
It does raise the issue as to whether /my_website/files/tmp/ directory should be replicated to individual machines or not. Just to make it more interesting the /my_website/files/cache directory is also changing randomly - as in directories and such coming and going.... If someone has helpful hints for actions to take post-rsync, please let me know For example: is it ok to stop services, delete everything in the /cache/ directory and restart? How about /files/tmp/ directory?
The mysql service will continue to be hosted on the network, so that's a non-issue.
After reading through how people handle "centralization", it looks like the /concrete in each web directory is meant to simply be kept at the same version, nothing more. There are no edits going on in that directory (one hopes!)
So, if all of my DocumentRoots are rsync copies of the master volume, and we carefully plan any C5 software updates, and edits to C5 sites are only done on a Master instance, then we should be in business, yes?
Suggestions/caveats/speedbumps/glaring errors are all welcome.
