Plone Cluster - Different physical machines

Hello friends,

i'm trying to put Plone to work in cluster. My idea is one physical machine to ZEO and two machines (physical) to clients.

Unfortunately, i'm having problems. First of all, the communication between clients/server is very good (tcpdump shows). But, when i try to put a site (addsite), after a time the error message appears:

IOError: [Errno 2] No such file or directory: '/usr/local/Plone/plonegov/var/blobstorage/0x00/0x00/0x00/0x00/0x00/0x00/0xa2/0x46/0x007TjZ3G.tmp-

Well, it is clear that plone is looking for blobstorage...but this folder would be in server, right? So, why looking for it in local directory? I searched this folder in local file, but i did'nt find. Same in the server... :frowning:

I followed this site as guide: https://docs.plone.org/4/en/old-reference-manuals/buildout/zope_to_zeo.html

I appreciate any help.

Thanks!

Fred

shared-blob = off

in your buildout configuration will do it.

-aj

1 Like

the standard configuration for a ZEO Cluster is with shared blobs, @zopyx is right with his answer that for you switching shared-blob = off will solve this. But tis also means higher load of the zeo-server and internal network traffic. It would be recommendend to put the blobs on a shared medium, a high avaliable network share or simply a nfs share of the zeo-server.

In detail ther is a difference between a shared blobstore and a blob-cache on zeo-clients. see https://pypi.python.org/pypi/plone.recipe.zope2instance#basic-zeo-storage

Transfering blobs over ZEO or over NFS...will not make much difference. You have to transfer the blobs in both cases. Caching strategies might make the differences. In general there should not br much difference with normal sites. A local blob cache with shared-blobs=off may also help.

-aj

in the past we used NFS and saw no gain at all, so we prefer to let ZEO take care of the blobs also; anyway the blobs will be locally cached in the client servers and that will reduce the network load.

Thanks a lot. Worked very well.

Fred