| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
| |
When creating a cluster with an addition AD server node, the DNS
forwarder on the AD node needs to be different to the DNS server for
the cluster nodes.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
| |
No use having them print the format string. That actually makes it
more restrictive. Instead, have the functions take cluster name and
index as arguments - these can then be passed to printf in either
order depending on the format string.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
This hasn't been used for a long time
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
The qcow2 + virtio + writeback problem seems to be fixed.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
| |
Some Samba packaging schemes have the winbind pieces in different
packages, so try to install those too. There's no check for failures,
so this is harmless (apart from warning messages) when the packages
don't exist.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
| |
Replace the hardcoded node name regexp with a generated one. Detailed
comments are required to describe the construction of the new regexp.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
... and just call it once before doing any cluster operations.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
| |
As more node types appear this becomes unwieldy so don't be as
paranoid.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Reduce the amount of work needed to define how to create a new node
type.
* Each node type definition needs to define node_name_format_<TYPE>()
* Node types that should be part of the CTDB cluster should define
node_is_ctdb_node_<TYPE>() and have it echo 1
* Nodes need not define create_node_<TYPE>() if they don't do anything
unusual.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
The backward compatibility hacking is horrible!
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Remove configuration variable CLUSTER_TYPE
* Add "build" node type
* Rename setup scripts for "ad" and "build" node type
* Separate out some setup_node_nas.sh from setup_cluster_nas.sh
* Make install_packages.sh and cluster_setup.sh more verbose
* Separate out clusterfs_setup.sh
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
Rewrite cluster_setup() to factor out the common SSH bits.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
Rename 50.sofs.defconf too.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
| |
Move the code to new a new file and globally rename variable
NSD_SERVERS to NODES_STORAGE_GPFS.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
| |
... into hack_nodes_sofs_front() and hack_nodes_sofs_storage(). This
allows 50.sofs.defconf to be split
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently this is distributed around the node logic and is hard to
manage.
Instead, introduce 2 new functions (well, meta-functions, called using
call_func):
* is_dedicated_storage_node()
Defaults to false. Is true for sofs_storage.
* node_has_shared_disks()
Defaults to false. True for sofs_storage. True for sofs_front when
there are no dedicated storage nodes.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
They are not part of the CTDB cluster and do not have shared disks.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
They have been unused since public address generation was changed.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
It is done later when nodes are created.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
| |
-S DISK is the default on at least RHEL 5 and 6, so that should
collect the disk statistics we want.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
Failure here doesn't matter, so add it unconditionally.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Remove the mknsd.sh and nsd_server_common.bash scripts, merging their
functionaility into the main script.
Separate the logic into functions, and default to calling each
function in turn. Arguments can be passed to potentially run only
some functions but this is probably pointless.
Clean up a lot of contorted logic for building the GPFS nodes file and
deciding on primary/secondary nodes. Create all files in the same
directory as the scripts (instead of /tmp) since they should be
available for post-mortem anyway.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Single entry point is now setup_cluster.sh.
* Rename old setup_cluster.sh to setup_cluster_nas.sh.
* Split setup_ad_server.sh into install_ad_server.sh and
setup_cluster_ad_server.sh.
* Split setup_build.sh into install_build.sh and
setup_cluster_build.sh.
* Rename setup_gpfs.sh to setup_clusterfs_gpfs.sh.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
... and remove functions file.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* New script install_packages.sh is the single entry-point for
installing packages outside kickstart.
* New configuration variable CONFIG_EXTRA_PACKAGES has space-separated
list of extra packages to install. This can help if dependencies
are broken.
* Rename install_gpfs_nas.sh to just install_nas.sh, now just installs
packages (e.g. samba, ctdb, rssh) to support protocols.
* Rename install_gpfs.sh to install_clusterfs_gpfs.sh to provide a
hook to support installation of alternate cluster filesystem
packages.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
It is now cluster-filesystem-agnostic.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
|
|
|
|
|
| |
* CLUSTERFS_DEFAULT_MOUNTPOINT replaces GPFS_DEFAULT_MOUNTPOINT
* Tweak various scripts and configuration files to allow cluster
filesystem type to be configured
* Also tweak chroot-related options to not have GPFS-specific values
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
This hasn't been used in a long time.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
"onnode -P" implements the only useful feature.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
|
|
| |
The functions are unnecessary.
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|
|
|
|
| |
Signed-off-by: Martin Schwenke <martin@meltin.net>
|