chunmei [Fri, 31 Mar 2023 02:49:33 +0000 (02:49 +0000)]
crimson/osd: enable PGShardMapping access from all cores
Previously, all accesses (incuding loookups) had to occur on core 0. Now that
we want to be able to dispatch from all cores, we need PGShardManager to be
accessible from all cores. To that end, we now proxy updates to core 0, but maintain
local copies of the map so that cores can perform local lookups.
Casey Bodley [Thu, 22 Jun 2023 12:47:06 +0000 (08:47 -0400)]
qa/s3tests: make extra_attrs additive
the s3tests.py task is filtering out several attrs by default. but
when dbstore uses `extra_attrs` to add 'not fails_on_dbstore', it
overwrites those other filters
John Mulligan [Mon, 19 Jun 2023 16:54:24 +0000 (12:54 -0400)]
cephadm: dont set ctx.image if json deploy image is unset
If no image has been provided in the "deploy from" json do not set
ctx.image to it (empty-string or None) as we may have had a valid
value passed on the --image CLI option.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Ilya Dryomov [Fri, 16 Jun 2023 12:01:52 +0000 (14:01 +0200)]
qa/workunits/rbd: make continuous export-diff test actually work
The current version is pretty useless:
- "rbd bench" writes the same byte (0xff) over and over again, so
almost all checksumming is in vain
- snapshots are taken in a steady state (i.e. not under I/O), so no
race conditions can get exposed
- even with these caveats, it's not wired up into the suite
Redo this workunit to be a reliable reproducer for the issue fixed
in the previous commit and wire it up for both krbd and rbd-nbd.
Venky Shankar [Tue, 20 Jun 2023 10:04:30 +0000 (15:34 +0530)]
Merge PR #51500 into main
* refs/pull/51500/head:
test/libcephfs: add test case for revoking caps
client: issue a cap release immediately if no cap exists
mds: add the revoking caps back to _revokes list
mds: move confirm_receipt() to Capability.cc
Jos Collin [Mon, 22 May 2023 04:31:39 +0000 (10:01 +0530)]
mds: display sane hex value (0x0) for empty feature bit
Print a valid hex (0x0) during empty feature bit, so that the clients could recognize it.
When the _vec size becomes 0, print() function creates an invalid hex (0x) and 'perf stats'
crashes with the below error:
"
File "/opt/ceph/src/pybind/mgr/stats/fs/perf_stats.py", line 177, in notify_cmd
metric_features = int(metadata[CLIENT_METADATA_KEY]["metric_spec"]["metric_flags"]["feature_bits"], 16)
ValueError: invalid literal for int() with base 16: '0x'
"
This patch creates a valid hex (0x0), when _vec size is 0.
Fixes: https://tracker.ceph.com/issues/59551 Signed-off-by: Jos Collin <jcollin@redhat.com>
Ilya Dryomov [Tue, 13 Jun 2023 11:36:02 +0000 (13:36 +0200)]
librbd: stop passing IOContext to image dispatch write methods
This is a major footgun since any value passed e.g. at the API layer
may be stale by the time we get to object dispatch. All callers are
passing the IOContext returned by get_data_io_context() for their
ImageCtx anyway, highlighting that the parameter is fictitious.
Only the read method can meaningfully take IOContext.
Ilya Dryomov [Mon, 12 Jun 2023 19:45:03 +0000 (21:45 +0200)]
librbd: use an up-to-date snap context when owning the exclusive lock
By effectively moving capturing of the snap context to the API layer,
commit 1d0a3b17f590 ("librbd: pass IOContext to image-extent IO
dispatch methods") introduced a nasty regression. The snap context can
be captured only after exclusive lock is safely held for the duration
of dealing with the image request and even then must be refreshed if
a snapshot creation request is accepted from a peer. This is needed to
ensure correctness of the object map in general and fast-diff states in
particular (OBJECT_EXISTS vs OBJECT_EXISTS_CLEAN) and object deltas
computed based off of them. Otherwise the object map that is forked
for the snapshot isn't guaranteed to accurately reflect the contents of
the snapshot when the snapshot is taken under I/O (as in disabling the
object map may lead to different results being returned for reads).
The regression affects mainly differential backup and snapshot-based
mirroring use cases with object-map and/or fast-diff enabled: since
some object deltas may be incomplete, the destination image may get
corrupted.
This commit represents a reasonable minimal fix: IOContext passed
through to ImageDispatch is effected only for reads and just gets
ignored for writes. The next commit cleans up further by undoing the
passing of IOContext through the image dispatch layers for writes.
John Mulligan [Thu, 15 Jun 2023 19:54:09 +0000 (15:54 -0400)]
cephadm: remove call to get_parm from fetch_configs
Stop using get_parm in fetch_configs. Doing so makes clear that
the two if-branches in fetch_configs are symmetric, in the
handling of custom_config_files and symmetric with the behavior
of fetch_custom_config_files.
It also reveals that get_parm only has one remaining caller making it
simpler to remove get_parm in the future.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Tue, 30 May 2023 20:23:13 +0000 (16:23 -0400)]
cephadm: combine deploy functions some more
During review it was pointed out that much of command_deploy and
command_deploy_from were still common. Combine the bulk of
command_deploy and command_deploy_from into a new _common_deploy.
The old _common_deploy is renamed to _dispatch_deploy but kept
as a separate function.
The reason for keeping _dispatch_deploy as a function is that
it limits the scope of variables. It helps knowing exactly what
is needed and what is not needed from _common_deploy.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Thu, 1 Jun 2023 17:27:37 +0000 (13:27 -0400)]
cephadm: drop deploy_arguments key from deploy config JSON
Drop the now unnecessary deploy_arguments subsection from the JSON
object that the new `_orch deploy` takes. The intermediate code that
used it in the ceph mgr module has been replaced with `params`.
We still want to share some code with the old deploy, especially the
default values for various things that get attached to `ctx`. In
some cases this is particularly important as not all code checks
that `ctx` has an attribute first. Create an `ArgumentFacade` type
to reuse the `add_argument` calls for deploy, and instead store the
default key-value pairs for `ctx`.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Sat, 20 May 2023 17:42:02 +0000 (13:42 -0400)]
cephadm: split command_deploy_from into mockable parts
Split up command_deploy_from so that the part that reads the JSON
from an input source is moved into read_configuration_source and
the part that applied the configuration data to the ctx is moved
into apply_deploy_config_to_ctx. This will make these sections
reusable and, more importantly, make it easier to mock/monkeypatch
these functions in tests.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Thu, 18 May 2023 20:14:24 +0000 (16:14 -0400)]
cephadm: make the old command_deploy deprecated with decorator
Add a deprecated command decorator. Decorate the old command_deploy with
it. By default it just logs if a deprecated command was used, but
custom builds of cephadm can set the NO_DEPRECATED var to True
and fail if a deprecated command is used.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Thu, 18 May 2023 19:10:16 +0000 (15:10 -0400)]
mgr/cephadm: add exchange types and use them for deploy
Add an exchange module for defining the "exchange protocol" that the
mgr module uses for communicating with the cephadm binary. The
exchange module classes use data descriptors to define supported
fields in the resulting data and will be serialized to JSON for
communicating with cephadm.
The cephadm binary does not use these types because it currently
doesn't import anything outside of the standard library. A suggested
future change would be to move 'exhange.py' to python-common somewhere
and have the module be imported by both the mgr and cephadm and used
for both serialization and deserialization.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Tue, 16 May 2023 17:49:28 +0000 (13:49 -0400)]
mgr/cephadm: block future use of extra_args
Convert `extra_args` into a read-only property so that
current code continues to work but anything that tries
to set any `extra_args` now or in the future will be
intentionally broken.
This is mainly to find issues when running unit and teuthology tests.
This patch could be dropped before the final merge if desired.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Tue, 16 May 2023 17:48:58 +0000 (13:48 -0400)]
mgr/cephadm: convert extra_args to daemon_params dict
In serve.py we were creating various command line arguments
(and their values) to pass to cephadm. Now that we support
a single unified JSON-based input we can stop passing
these as CLI parameters and just pass it as a singular
JSONified dict.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Thu, 11 May 2023 15:07:53 +0000 (11:07 -0400)]
mgr/cephadm: deploy using deploy-from command
The new `deploy-from` command aims to be a cleaner, clearer way to pass
input to cephadm for service deployment. Make a minimal set of change to
use that new command.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Wed, 24 May 2023 18:11:52 +0000 (14:11 -0400)]
mgr/cephadm: allow passing a list as a compound command to _run_cephadm
As we're adding "namespaces" to the cephadm command it's easier to work
with a compound list of terms as the command. This also avoids an issue
where the function call injects args between the command and the "args"
passed to _run_cephadm.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Thu, 15 Jun 2023 19:34:15 +0000 (15:34 -0400)]
cephadm: update code to use fetch_custom_config_files
Update the code to use fetch_custom_config_files rather than
get_custom_config_files. Like other fetch_ functions recently added
to cephadm we don't LBYL for checking values on ctx, but rather
call the fetch function on ctx and then check if the result contains
actionable data.
get_custom_config_files gets removed as no remaining callers exist.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Thu, 15 Jun 2023 19:34:09 +0000 (15:34 -0400)]
cephadm: add fetch_custom_config_files function
This function works similarly to get_custom_config_files but doesn't
require the use of "ctx.config_json" and will prefer sourcing the
data from `config_blobs`.
This function also returns the list of dicts rather than return a dict
with exactly one key which maps to said list of dicts. There appeared
to be no benefit to that when the call sites of get_custom_config_files
was examined.
This function is `fetch_configs` evil twin.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Fri, 26 May 2023 19:06:40 +0000 (15:06 -0400)]
cephadm: add fetch_tcp_ports function for getting processed ports
The cephadm cli can pass ports as either a string containing space
separated port numbers or, in the newer method, a (JSON) list of ints.
The fetch_tcp_ports function abstracts away the handling of
these input forms.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan [Thu, 11 May 2023 15:05:07 +0000 (11:05 -0400)]
cephadm: add deploy-from command
Add the `deploy-from` command. This command is like `deploy` in
functionality but instead of sourcing input from a mixture of various
CLI arguments, CLI arguments that take JSON strings, and JSON data
provided on stdin - this command aims to take configuration from a
single unified JSON object provided on the commands stdin.
This change strives to be as reasonably backwards compatible with
`deploy` as possible and this leads to some awkward hoop-jumping,
including: needing to nest CLI arguments inside the JSON, because
that is what the manager constructs; duplicate sources of the fsid
(CLI and JSON); the strange behavior of CephadmContext when setting
values; and a few others I'm probably forgetting.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
projects / ceph.git / log