CLI: Add force option when creating a symlink. This will allow to overwrite an already defined symlink

Closes EOS-6392

different from 1MB. EOS-6383

layout of the source file. Fixes EOS-6383

This happens regularly, and also can be easily checked via the status commands,
do not log at WARN level

Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

Make the function that selects files allow a filter argument which defaults to
Null so that GroupBalancer can filter out proc files whereas the GroupDrainer
needs to not skip these.

Fixes: EOS-6382
Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

the request to the MGM that can handle it. This still requires the XrdCl
to support redirection for this request as currently 5.7.3 fails with the
following message:

[2025-03-07 16:20:19.947805 +0100][Debug  ][ExDbgMsg] Calling MsgHandler: 0x5590592e40d0 (message: kXR_unknown (length: 1) ) with status: [ERROR] Unhandled redirect.

In the case some paths are mapped to another EOS/XRootD instance (eos route ls),
a client performing a WEBDAV request against those paths will be redirected.

Fixes EOS-6370

A new interface is introduced which allows for fetching groups via
getgrouplist insted of getgrent which are rentrant. A env allows for this to be
configurable at runtime

The defaults have changed to getting grouplist, which means for a given user we
only fetch their groups with this, drastically reducing the calls to a stateful
getgrent and co which might be heavy and cause errors with lots of queries on
LDAP/SSSD backends.

Fixes: EOS-6356
Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

Can be ran on a machine that has nftables enabled. Pass a list of ports via --ports and the script
will flush the nft rules and create rules to specifically block the
incoming traffic towards those ports.

The changes are not persisted. Therefore any call to
systemctl restart nftables or any call to
eos-ports-block --reset-defaults will put back the previously configured
rules.

Fixes EOS-6368

Added corresponding system tests
Fixes EOS-6364

Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

conversion policy. Most of the time, these can be completely skipped if
the corresponding extended attribute is not present.

- skip collecting ns stat in these cases
- s/DUMP_DIR_NAME/DUMP_DIR in line with other conventions

Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

Fixes: EOS-6357
Reported-by: David Smith <David.Smith@cern.ch>
Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

This test will crash before fixes for EOS-6357 due to the unsafe access of
underlying async notifications vector

Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

With the default 20 placement attempts EC configurations might fail with
extremely small groups with weighted placements as weights might drop to 0
during an iteration, also random engines might pick the same FSid again burning
an attempt. Since running the loop is extremely cheap and cost of failed result
is very high, increase placement attempts.
Add tests with n_replica=12 to simulate EC layout behaviour which fail with low
attempt configuration

Fixes: EOS-6355

Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

Fixes: EOS-6349
Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

Signed-off-by: Abhishek Lekshmanan <abhishek.lekshmanan@cern.ch>

opened in the first place. These will be anyway silently recovered from
the other stripes.