IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
There is no reason to serialise these or even handle remote nodes
first. Using a broadcast is more efficient and is less code.
Update expected test results to reflect changed order of messages.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Autobuild-User(master): Amitay Isaacs <amitay@samba.org>
Autobuild-Date(master): Mon Mar 23 15:04:00 CET 2015 on sn-devel-104
"ctdb reloadnodes" currently does no sanity checking of the nodes
file. This can cause chaos if a line is deleted from the nodes file
rather than commented out. It also repeatedly produces a spurious
warning for each deleted node, even if the node was deleted a long
time ago.
Instead compare the nodemap with the contents of the local nodes file
to sanity check before attempting any reloads. Note that this is
still imperfect if the nodes files are inconsistent across nodes but
it is better. Also ensure that any nodes that are to be deleted are
already disconnected. Avoid trying to talk to deleted nodes.
The current implementation is a bit unfortunate when it comes to
deleting nodes. The most obvious alternative to the above complexity
would be to reloadnodes on the specified node first, then fetch the
node map (in which newly deleted nodes would be marked as such) and
then handle the remote nodes. However, the implementation of
reloadnodes is asynchronous and it only actions the reload after 1
second. This is presumably to avoid the recovery master noticing the
inconsistency between nodemaps and triggering a recovery before all
nodes have had their nodemaps updated.
Note that this recovery can still occur if the check is done at an
inconvenient time. A better long term approach might be to quiesce
the recovery master checks while reloadnodes is in progress.
Update a unit test to reflect the change.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
A basic test and some for cross-node consistency checking.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This compares the nodes file on the current node with that on all
nodes. If any are different then do not reload nodes.
If any nodes files can't be fetched then do not reload nodes. This
could be because some nodes are running an older version without this
feature. This is unsupported: why make a major cluster
reconfiguration while a cluster is half upgraded?
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This is like CTDB_CONTROL_GET_NODEMAP but it loads from the nodes file
instead of the daemon.
Also new client function ctdb_ctrl_getnodesfile()
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
It should not be possible to specify "-n <othernode>", unless
<othernode> is the current node. To support this, add new function
assert_current_node_only().
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
New function ctdb_read_nodes_file() reads a nodes file into a node
map, which is a useful intermediate format. This function should
replace the node reading code in the ctdb CLI tool. It will also be
useful for sanity checking of nodes files across the cluster.
New function convert_node_map_to_list() converts a node map to a node
array (and associated node count). This fills in the details that
aren't present in the node map. This may also useful as a separate
function later if node list reloading stages the data after a sanity
check - the approach is not yet finalised.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Every time a nodemap is contructed the node IP addresses all need to
be parsed. This isn't very productive use of CPU.
Instead, parse each string once when the nodes file is loaded. This
results in much simpler code.
This code also removes the use of ctdb_address. Duplicating the port
is pointless without an abstraction layer around ctdb_address. If
CTDB gets an incompatible transport in the future then add an
abstraction layer.
Note that the infiniband code is not updated. Compilation of the
infiniband code is already broken. Fixing it will be a separate,
properly tested effort.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Pair-programmed-with: Amitay Isaacs <amitay@gmail.com>
Having it require a CTDB context stops ctdb_parse_address() from being
used in more generic code. Just use the existing talloc context for
memory allocations.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Just add a flags parameter to ctdb_add_nodes() and use the same code.
Less is more.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
This is currently set in 2 places. One of them makes the node loading
code difficult to refactor. Also, when the surrounding code in either
place is touched then it might get broken.
This only needs to be done once at startup, not on every reload. So
do it once in a very obvious way, sacrificing a few CPU cycles for
some added clarity.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Each node reload unnecessarily and incorrectly resets the VNN map,
causing a potentially unnecessary recovery. When nodes are reloaded
any newly deleted nodes should already be disconnected and any newly
added nodes should also be disconnected. This means that reloading
the nodes file should not cause a change in the VNN map.
The current implementation also leaks memory every time the nodes are
reloaded.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
It should be -1 even without a failure callback registered.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
These can be unset if a NODEMAP, IFACES or VNNMAP section is missing.
Affected functions would then dereference a NULL pointer and the test
program would crash. Adding some helpful messages makes the problem
easier to diagnose when writing tests.
Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
Pair-Programmed-With: Guenther Deschner <gd@samba.org>
Pair-Programmed-With: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Andreas Schneider <asn@samba.org>
Signed-off-by: Guenther Deschner <gd@samba.org>
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Autobuild-User(master): Stefan Metzmacher <metze@samba.org>
Autobuild-Date(master): Sat Mar 21 02:03:34 CET 2015 on sn-devel-104
