previously, it was a requirement to have attacher sidecar for CSI
drivers and there had an implementation of dummy mode of operation.
However skipAttach implementation has been stabilized and the dummy
mode of operation is going to be removed from the external-attacher.
Considering this driver work on volumeattachment objects for NBD driver
use cases, we have to implement dummy controllerpublish and unpublish
and thus keep supporting our operations even in absence of dummy mode
of operation in the sidecar.
This commit make a NOOP controller publish and unpublish for RBD driver.
CephFS driver does not require attacher and it has already been made free
from the attachment operations.
Ref# https://github.com/ceph/ceph-csi/pull/3149
Ref# https://github.com/kubernetes-csi/external-attacher/issues/226
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
After Failover of workloads to the secondary
cluster when the primary cluster is down,
RBD Image is not marked healthy, and VR
resources are not promoted to the Primary,
In VolumeReplication, the `CURRENT STATE`
remains Unknown and doesn't change to Primary.
This happens because the primary cluster went down,
and we have force promoted the image on the
secondary cluster. and the image stays in
up+stopping_replay or could be any other states.
Currently assumption was that the image will
always be `up+stopped`. But the image will be in
`up+stopped` only for planned failover and it
could be in any other state if its a forced
failover. For this reason, removing
checkHealthyPrimary from the PromoteVolume RPC call.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Recently the k8s.io/mount-utils package added more runtime dectection.
When creating a new Mounter, the detect is run every time. This is
unfortunate, as it logs a message like the following:
```
mount_linux.go:283] Detected umount with safe 'not mounted' behavior
```
This message might be useful, so it probably good to keep it.
In Ceph-CSI there are various locations where Mounter instances are
created. Moving that to the DefaultNodeServer type reduces it to a
single place. Some utility functions need to accept the additional
parameter too, so that has been modified as well.
See-also: kubernetes/kubernetes#109676
Signed-off-by: Niels de Vos <ndevos@redhat.com>
When the Ceph user is restricted to a specific namespace in the pool, it is
crucial that evey interaction with the cluster is done within that namespace.
This wasn't the case in `getCloneDepth()`.
This issue was causing snapshot creation to fail with
> Failed to check and update snapshot content: failed to take snapshot of the
> volume X: "rpc error: code = Internal desc = rbd: ret=-1, Operation not
> permitted"
Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>
There are regular reports that identify a non-error as the cause of
failures. The Kubernetes mount-utils package has detection for systemd
based environments, and if systemd is unavailable, the following error
is logged:
Cannot run systemd-run, assuming non-systemd OS
systemd-run output: System has not been booted with systemd as init
system (PID 1). Can't operate.
Failed to create bus connection: Host is down, failed with: exit status 1
Because of the `failed` and `exit status 1` error message, users might
assume that the mounting failed. This does not need to be the case. The
container-images that the Ceph-CSI projects provides, do not use
systemd, so the error will get logged with each mount attempt.
By using the newer MountSensitiveWithoutSystemd() function from the
mount-utils package where we can, the number of confusing logs get
reduced.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
This commit removes the clone incase
unsetAllMetadata or copyEncryptionConfig or
expand fails for createVolumeFromSnapshot
and CreateSnapshot.
It also removes the clone in case of
any failure in createCloneFromImage.
issue: #3103
Signed-off-by: Yati Padia <ypadia@redhat.com>
As we added support to set the metadata on the rbd images created for
the PVC and volume snapshot, by default metadata is set on all the images.
As we have seen we are hitting issues#2327 a lot of times with this,
we start to leave a lot of stale images. Currently, we rely on
`--extra-create-metadata=true` to decide to set the metadata or not,
we cannot set this option to false to disable setting metadata because we
use this for encryption too.
This changes is to provide an option to disable setting the image
metadata when starting cephcsi.
Fixes: #3009
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Removed the code in checkHealthyPrimary which
makes the ceph call, passing it as input now.
Added unit test for checkHealthyPrimary function
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
we need to check for image should be in up+stopped state
not anyone of the state for that the we need to use
OR check not the AND check.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
When the image is force promoted to primary on the
cluster the remote image might not be in replaying
state because due to the split brain state. This
PR reverts back the commit
c3c87f2ef3. Which we added
to check the remote image status.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Kubernetes 1.24 and newer use a different path for staging the volume.
That means the CSI-driver is requested to mount the volume at an other
location, compared to previous versions of Kubernetes. CSI-drivers
implementing the volumeHealer, must receive the correct path, otherwise
the after a nodeplugin restart the NBD mounts will bailout attempting
to NodeStageVolume() call and return an error.
See-also: kubernetes/kubernetes#107065
Fixes: #3176
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
During failover we do demote the volume on the primary
as the image is still not promoted yet on the remote cluster,
there are spurious split-brain errors reported by RBD,
the Cephcsi resync will attempt to resync from the "known"
secondary and that will cause data loss
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
RBD supports creating rbd images with
object size, stripe unit and stripe count
to support striping. This PR adds the support
for the same.
More details about striping at
https://docs.ceph.com/en/quincy/man/8/rbd/#stripingfixes: #3124
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
This change helps read the cluster name from the cmdline args,
the provisioner will set the same on the RBD images.
Fixes: #2973
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
continue running rbd driver when /sys/bus/rbd/supported_features file is
missing, do not bailout.
Fixes: #2678
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
krbdFeatures is set to zero when kernel version < 3.8, i.e. in case where
/sys/bus/rbd/supported_features is absent and we are unable to prepare
the krbd attributes based on kernel version.
When krbdFeatures is set to zero fallback to NBD only when autofallback
is turned ON.
Fixes: #2678
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Upstream /sys/bus/rbd/supported_features is part of Linux kernel v4.11.0
Prepare the attributes and use them in case if
/sys/bus/rbd/supported_features is missing.
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Move k8s.GetVolumeMetadata() out of setVolumeMetadata() and rename it to
setAllMetadata() so that the same can be used for setting volume and
snapshot metadata.
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
For the default mounter the mounter option
will not be set in the storageclass and as it is
not available in the storageclass same will not
be set in the volume context, Because of this the
mapOptions are getting discarded. If the mounter
is not set assuming it's an rbd mounter.
Note:- If the mounter is not set in the storageclass
we can set it in the volume context explicitly,
Doing this check-in node server to support backward
existing volumes and the check is minimal we are not
altering the volume context.
fixes: #3076
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
This commit makes modification so as to allow pvc-pvc clone
with different storageclass having different encryption
configs.
This commit also modifies `copyEncryptionConfig()` to
include a `isEncrypted()` check within the function.
Signed-off-by: Rakshith R <rar@redhat.com>
As the netNamespaceFilePath can be separate for
both cephfs and rbd adding the netNamespaceFilePath
path for RBD, This will help us to keep RBD and
CephFS specific options separately.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
To consider the image is healthy during the Promote
operation currently we are checking only the image
state on the primary site. If the network is flaky
or the remote site is down the image health is
not as expected. To make sure the image is healthy
across the clusters check the state on both local
and the remote clusters.
some details:
https://bugzilla.redhat.com/show_bug.cgi?id=2014495
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
calling setRbdNbdToolFeatures inside an init
gets called in main.go for both cephfs and rbd
driver. instead of calling it in init function
calling this in rbd driver.go as this is specific
to rbd.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Set snapshot-name/snapshot-namespace/snapshotcontent-name details
on RBD backend snapshot image as metadata on snapshot
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Make sure to set metadata when image exist, i.e. if the provisioner pod
is restarted while createVolume is in progress, say it created the image
but didn't yet set the metadata.
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Example if a PVC was delete by setting `persistentVolumeReclaimPolicy` as
`Retain` on PV, and PV is reattached to a new PVC, we make sure to update
PV/PVC image metadata on a PV reattach.
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
This helps Monitoring solutions without access to Kubernetes clusters to
display the details of the PV/PVC/NameSpace in their dashboard.
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Define and use PV and PVC metadata keys used by external provisioner.
The CSI external-provisioner (v1.6.0+) introduces the
--extra-create-metadata flag, which automatically sets map<string, string>
parameters in the CSI CreateVolumeRequest.
Add utility functions to set/Get PV/PVC/PVCNamespace metadata on image
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
add support to run rbd map and mount -t
commands with the nsenter.
complete design of pod/multus network
is added here https://github.com/rook/rook/
blob/master/design/ceph/multus-network.md#csi-pods
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Currently we only check if the rbd-nbd tool supports cookie feature.
This change will also defend cookie addition based on kernel version
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Increase the timeout to 2 minutes to give enough time
for rollback to complete.
As rollback is performed by the force-promote command it,
at times, may take more than a minute
(based on dirty blocks that need to be rolled
back approximately) to rollback.
The added extra 1 minute is useful though to avoid
multiple calls to complete the rollback and in
extremely corner cases to avoid failures in the
first instance of the call when the mirror watcher
is not yet removed (post scaling down the
RBD mirror instance)
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Restoring a snapshot with a new PVC results with a wrong
dataPoolName in case of initial volume linked
to a storageClass with topology constraints and erasure coding.
Signed-off-by: Thibaut Blanchard <thibaut.blanchard@gmail.com>
Below are the 3 different cases where we need
the PVC namespace for encryption
* CreateVolume:- Read the namespace from the
createVolume parameters and store it in the omap
* NodeStage:- Read the namespace from the omap
not from the volumeContext
* Regenerate:- Read the pvc namespace from the claimRef
not from the volumeAttributes.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
remove kubernetes csi prefixed parameters
from the volumeContext as we dont want
to store it in the PV VolumeAttributes.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
This commit ensures that parent image is flattened before
creating volume.
- If the data source is a PVC, the underlying image's parent
is flattened(which would be a temp clone or snapshot).
hard & soft limit is reduced by 2 to account for depth that
will be added by temp & final clone.
- If the data source is a Snapshot, the underlying image is
itself flattened.
hard & soft limit is reduced by 1 to account for depth that
will be added by the clone which will be restored from the
snapshot.
Flattening step for resulting PVC image restored from snapshot is removed.
Flattening step for temp clone & final image is removed when pvc clone is
being created.
Fixes: #2190
Signed-off-by: Rakshith R <rar@redhat.com>
Currently, we support
mapOption: "krbd:v1,v2,v3;nbd:v1,v2,v3"
- By omitting `krbd:` or `nbd:`, the option(s) apply to
rbdDefaultMounter which is krbd.
- A user can _override_ the options for a mounter by specifying `krbd:`
or `nbd:`.
mapOption: "v1,v2,v3;nbd:v1,v2,v3"
is effectively the same as the 1st example.
- Sections are split by `;`.
- If users want to specify common options for both `krbd` and `nbd`,
they should mention them twice.
But in case if the krbd or nbd specifc options contian `:` within them,
then the parsing is failing now.
E0301 10:19:13.615111 7348 utils.go:200] ID: 63 Req-ID:
0001-0009-rook-ceph-0000000000000001-fd37c41b-9948-11ec-ad32-0242ac110004
GRPC error: badly formatted map/unmap options:
"krbd:read_from_replica=localize,crush_location=zone:zone1;"
This patch fix the above case where the options itself contain `:`
delimitor
ex: krbd:v1,v2,v3=v31:v32;nbd:v1,v2,v3"
Please note, if you are using such options which contain `:` delimiter,
then it is mandatory to specify the mounter-type.
Fixes: #2910
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Makes the rbd images features in the storageclass
as optional so that default image features of librbd
can be used. and also kept the option to user
to specify the image features in the storageclass.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
as deep-flatten is long supported in ceph and its
enabled by default in the librbd, providing an option
to enable it in cephcsi for the rbd images we are
creating.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
logging the error is not user-friendly and
it contains system error message. Log the
stderr which is user-friendly error message
for identifying the problem.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Return the dataPool used to create the image instead of the default one
provided by the createVolumeRequest.
In case of topologyConstrainedDataPools, they may differ.
Don't add datapool if it's not present
Signed-off-by: Sébastien Bernard <sebastien.bernard@sfr.com>
This commit removes the thick provisioning
code as thick provisioning is deprecated in
cephcsi 3.5.0.
fixes: #2795
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
currently we are overriding the permission to `0o777` at time of node
stage which is not the correct action. That said, this permission
change causes an extra permission correction at time of nodestaging
by the CO while the FSGROUP change policy has been set to
`OnRootMismatch`.
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
as ioutil.ReadFile is deprecated and
suggestion is to use os.ReadFile as
per https://pkg.go.dev/io/ioutil updating
the same.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
as ioutil.WriteFile is deprecated and
suggestion is to use os.WriteFile as
per https://pkg.go.dev/io/ioutil updating
the same.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Currently most of the internal methods have the
rbdVolume as the received. As these methods
are completely internal and requires only
the fields of the rbdImage use rbdImage
as the receiver instead of rbdVolume.
updates #2742
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
as per the CSI standard the size is optional parameter,
as we are allowing the clone to a bigger size
today we need to block the clone to a smaller size
as its a have side effects like data corruption etc.
Note:- Even though this check is present in kubernetes
sidecar as CSI is CO independent adding the check
here.
updates: #2718
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
as per the CSI standard the size is optional parameter,
as we are allowing the restore to a bigger size
today we need to block the restore to a smaller size
as its a have side effects like data corruption.
Note:- Even though this check is present in kubernetes
sidecar as CSI is CO independent adding the check
here.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
SINGLE_NODE_WRITER capability ambiguity has been fixed in csi spec v1.5
which allows the SP drivers to declare more granular WRITE capability in form
of SINGLE_NODE_SINGLE_WRITER or SINGLE_NODE_MULTI_WRITER.
These are not really new capabilities rather capabilities introduced to
get the desired functionality from CO side based on the capabilities SP
driver support for various CSI operations, this new capabilities also help
to address new access mode RWOP (readwriteoncepod).
This commit adds a helper function which identity the request is of
multiwriter mode and also validates whether it is filesystem mode or
block mode. Based on the inspection it fails to allow multi write
requests for filesystem mode and only allow multi write request against
block mode.
This commit also adds unit tests for isMultiWriterBlock function which
validates various accesstypes and accessmodes.
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
This commit removes rbdVol.getTrashPath() function
since it is no longer being used due to introduction
of go-ceph rbd admin task api for deletion.
Signed-off-by: Rakshith R <rar@redhat.com>
With introduction of go-ceph rbd admin task api, credentials are
no longer required to be passed as cli cmd is not invoked.
Signed-off-by: Rakshith R <rar@redhat.com>
This commit removes `rv.Connect(cr)` since the rbdVolume should
have an active connection in this stage of the function call.
`rv.getCloneDepth(ctx)` will work after a connect to the cluster.
Signed-off-by: Rakshith R <rar@redhat.com>
This commit adds support to go-ceph rbd task api
`trash remove` and `flatten` instead of using cli
cmds.
Fixes: #2186
Signed-off-by: Rakshith R <rar@redhat.com>
The CSI Controller (provisioner) can call `rbd sparsify` to reduce the
space consumption of the volume.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
use ExecCommandWithTimeout with timeout
of 1 minute for the promote operation.
If the command doesnot returns error/response
in 1 minute the process will be killed
and error will be returned to the user.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
after creating the rbd image log the image
details corresponding for the request along
with the request name.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
as getImageInfo is already called inside
cloneRbdImageFromSnapshot function right
after creating the clone. remove the extra
API call to get the details again.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
after creating the clone get the current
image details like size, creationTime,
imageFeatures etc from the ceph cluster.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
moved ParentName, ParentPool and ImageFeatureSet
fields to the rbdImage struct as these are the
first citizens on the rbdImage.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
If the volume with a bigger size is created
from a snapshot or from another volume we
need to exapand the filesystem also in the
csidriver as nodeExpand request is not triggered
for this one, During NodeStageVolume we can
expand the filesystem by checking filesystem
needs expansion or not.
If its a encrypted device, check the device
size of rbd device and the LUKS device if required
the device will be expanded before
expanding the filesystem.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
If the requested volume size is greater than
the snapshot size, resize the cloned volume
after creating a clone from a snapshot.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
If the requested volume size is greater than
the parent volume size, resize the cloned volume
after creating a final clone from a parent volume.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
added a check to consider ErrImageNotFound error
during DeleteSnapshot operation, if the error
is ErrImageNotFound we need to ensure that image
is removed from the trash and also the rados
OMAP data is removed.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
we need actual size of the rbdVolume
created for the snapshot, as we are not
storing the size of the snapshot in OMAP
we need to fetch the size from ceph cluster
and update the same on rbdSnapshot struct.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
as we are moving the VolSize to rbdImage struct
we should reuse the same instead of maintaining
one more field in rbdSnapshot struct.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
move the Volsize to the rbdImage struct
as size is more applicable for rbdImage
as rbdImage is used for both rbdVolume
and rbdSnapshot.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
as we are no longer supporting the v1.x
version of cephcsi. removing the json tag
used to store rbd volume details in configmap.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
when doing the internal operation to get the
latest details the rbd image size is also getting
updated and this will update the volume size also
without actual requested size we cannot do the
resize operation for bigger clones. This commit
adds a new field called RequestedVolSize to rbdVolume
struct to hold the user requested size.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
added a new helper function called cleanupThickClone
to cleanup the snapshot and clone if the thick
provisioning is not fully completed.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
remove the bigger size validation when
creating a volume from a snapshot or when
creation a clone from a volume as we resized
the volume after cloning.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
dummy image rbdVolume struct is derived
from the actual one rbdVolume of the
volumeID sent in the EnableVolumeReplication
request. and the dummy rbdVolume struct contains
the image id of the actual volume because
of that when we are repairing the dummy
image the image is sent to trash but not
deleted due to the wrong image ID. resetting
the image id will makes sure the image id
is fetching from ceph cluster and same
image id will be used for manager operation.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
IsBlockMultiNode() is a new helper that takes a slice of
VolumeCapability objects and checks if it includes multi-node access
and/or block-mode support.
This can then easily be used in other services that need checking for
these particular capabilities, and preventing multi-node block-mode
access.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
we have added clusterID mapping to identify the volumes
in case of a failover in Disaster recovery in #1946.
with #2314 we are moving to a configuration in
configmap for clusterID and poolID mapping.
and with #2314 we have all the required information
to identify the image mappings.
This commit removes the workaround implementation done
in #1946.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
The rbd package contains several functions that can be used by
CSI-Addons Service implmentations. Unfortunately it is not possible to
do this, as the rbd-driver needs to import the csi-addons/rbd package to
provide the CSI-Addons server. This causes a circular import when
services use the rbd package:
- rbd/driver.go import csi-addons/rbd
- csi-addons/rbd import rbd (including the driver)
By moving rbd/driver.go into its own package, the circular import can be
prevented.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
HexStringToInteger() used to return a uint64, but everywhere else uint
is used. Having HexStringToInteger() return a uint as well makes it a
little easier to use when setting it with SetGlobalInt().
Signed-off-by: Niels de Vos <ndevos@redhat.com>
When the rbd-driver starts, it initializes some global (yuck!) variables
in the rbd package. Because the rbd-driver is moved out into its own
package, these variables can not easily be set anymore.
Introcude SetGlobalInt(), SetGlobalBool() and InitJournals() so that the
rbd-driver can configure the rbd package.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
The rbd-driver calls rbd.runVolumeHealer() which is not available
outside the rbd package. By moving the rbd-driver into its own package,
RunVolumeHealer() needs to be exported.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
NodeServer.mounter is internal to the NodeServer type, but it needs to
be initialized by the rbd-driver. The rbd-driver is moved to its own
package, so .Mounter needs to be available from there in order to set
it.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
genVolFromVolID() is used by the CSI Controller service to create an
rbdVolume object from a CSI volume_id. This function is useful for
CSI-Addons Services as well, so rename it to GenVolFromVolID().
Signed-off-by: Niels de Vos <ndevos@redhat.com>
The dummy image will be created with 1Mib size.
during the snapshot transfer operation the 1Mib
will be transferred even if the dummy image doesnot
contains any data. adding the new image features
`fast-diff,layering,obj-map,exclusive-lock`on the
dummy image will ensure that only the diff is
transferred to the remote cluster.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
we added a workaround for rbd scheduling by creating
a dummy image in #2656. with the fix we are creating
a dummy image of the size of the first actual rbd
image which is sent in EnableVolumeReplication request
if the actual rbd image size is 1TiB we are creating
a dummy image of 1TiB which is not good. even though
its a thin provisioned rbd images this is causing
issue for the transfer of the snapshot during
the mirroring operation.
This commit recreates the rbd image with 1MiB size
which is the smaller supported size in rbd.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
rbd mirroring CLI calls are async and it doesn't wait
for the operation to be completed. ex:- `rbd mirror image enable`
it will enable the mirroring on the image but it doesn't
ensure that the image is mirroring enabled and healthy
primary. The same goes for the promote volume also.
This commits adds a check-in PromoteVolume to make sure
the image in a healthy state i.e `up+stopped`.
note:- not considering any intermediate states to make
sure the image is completely healthy before responding
success to the RPC call.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Journal-based RADOS block device mirroring ensures point-in-time
consistent replicas of all changes to an image, including reads and
writes, block device resizing, snapshots, clones, and flattening.
Journaling-based mirroring records all modifications to an image in the
order in which they occur. This ensures that a crash-consistent mirror
of an image is available.
Mirroring when configured in journal mode, mirroring will
utilize the RBD journaling image feature to replicate the image
contents. If the RBD journaling image feature is not yet enabled on the
image, it will be automatically enabled.
Fixes: #2018
Co-authored-by: Madhu Rajanna <madhupr007@gmail.com>
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Depending on the way Ceph-CSI is deployed, the capabilities will be
configured for the GetCapabilities procedure. The other procedures are
more straight-forward.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
After adding the new CSI-Addons Server, golang-ci complains that
driver.Run() is too complex. By moving the profiling checks and starting
of the go-routines in their own function, golang-ci is happy again.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
currently we are fist operating on the dummy
image to refresh the pool and then we are adding
the scheduling. we think the scheduling should
be added first and than we should refresh the
pool. If we do this all the existing schedules
will be considered from the scheduler.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
with shallow copy of rbdVol to dummyVol
the image name update of the dummyVol is getting
reflected on the rbdVol which we dont want.
do deep copy to avoid this problem.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Uses the below schema to supply mounter specific map/unmapOptions to the
nodeplugin based on the discussion we all had at
https://github.com/ceph/ceph-csi/pull/2636
This should specifically be really helpful with the `tryOthermonters`
set to true, i.e with fallback mechanism settings turned ON.
mapOption: "kbrd:v1,v2,v3;nbd:v1,v2,v3"
- By omitting `krbd:` or `nbd:`, the option(s) apply to
rbdDefaultMounter which is krbd.
- A user can _override_ the options for a mounter by specifying `krbd:`
or `nbd:`.
mapOption: "v1,v2,v3;nbd:v1,v2,v3"
is effectively the same as the 1st example.
- Sections are split by `;`.
- If users want to specify common options for both `krbd` and `nbd`,
they should mention them twice.
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
The dummy mirror image needs to be disabled and then
reenabled for mirroring, to ensure a newly promoted
primary is now starting to schedule snapshots.
Signed-off-by: Shyamsundar Ranganathan <srangana@redhat.com>
currently we have a bug in rbd mirror scheduling module.
After doing failover and failback the scheduling is not
getting updated and the mirroring snapshots are not
getting created periodically as per the scheduling
interval. This PR workarounds this one by doing below
operations
* Create a dummy (unique) image per cluster and this image
should be easily identified.
* During Promote operation on any image enable the
mirroring on the dummy image. when we enable the mirroring
on the dummy image the pool will get updated and the
scheduling will be reconfigured.
* During Demote operation on any image disable the mirroring
on the dummy image. the disable need to be done to enable
the mirroring again when we get the promote request to make
the image as primary
* When the DR is no more needed, this image need to be
manually cleanup as for now as we dont want to add a check
in the existing DeleteVolume code path for delete dummy image
as it impact the performance of the DeleteVolume workflow.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Moved to add scheduling to the promote
operation as scheduling need to be added
when the image is promoted and this is
the correct method of adding the scheduling
to make the scheduling take place.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
instead of logging the volumeID and the pool
name. log the poolname and image name for better
debugging.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
On line 341 a `transaction` is created. This is passed to the deferred
`undoStagingTransaction()` function when an error in the
`NodeStageVolume` procedure is detected. So far, so good.
However, on line 356 a new `transaction` is returned. This new
`transaction` is not used for the defer call.
By removing the empty `transaction` that is used in the defer call, and
calling `undoStagingTransaction()` on an error of `stageTransaction()`,
the code is a little simpler, and the cleanup of the transaction should
be done correctly now.
Updates: #2610
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Problem:
On remap/attach of device (i.e. nodeplugin restart), there is no way
for rbd-nbd to defend if the backend storage is matching with the initial
backend storage.
Say, if an initial map request for backend "pool1/image1" got mapped to
/dev/nbd0 and the userspace process is terminated (on nodeplugin restart).
A next remap/attach (nodeplugin start) request within reattach-timeout is
allowed to use /dev/nbd0 for a different backend "pool1/image2"
For example, an operation like below could be dangerous:
$ sudo rbd-nbd map --try-netlink rbd-pool/ext4-image
/dev/nbd0
$ sudo blkid /dev/nbd0
/dev/nbd0: UUID="bfc444b4-64b1-418f-8b36-6e0d170cfc04" TYPE="ext4"
$ sudo pkill -15 rbd-nbd <-- nodeplugin terminate
$ sudo rbd-nbd attach --try-netlink --device /dev/nbd0 rbd-pool/xfs-image
/dev/nbd0
$ sudo blkid /dev/nbd0
/dev/nbd0: UUID="d29bf343-6570-4069-a9ea-2fa156ced908" TYPE="xfs"
Solution:
rbd-nbd/kernel now provides a way to keep some metadata in sysfs to identify
between the device and the backend, so that when a remap/attach request is
made, rbd-nbd can compare and avoid such dangerous operations.
With the provided solution, as part of the initial map request, backend
cookie (ceph-csi VOLID) can be stored in the sysfs per device config, so
that on a remap/attach request rbd-nbd will check and validate if the
backend per device cookie matches with the initial map backend with the help
of cookie.
At Ceph-csi we use VOLID as device cookie, which will be unique, we pass
the VOLID as cookie at map and use the same at the time of attach, that
way rbd-nbd can identify backends and their matching devices.
Requires:
https://github.com/ceph/ceph/pull/41323https://lkml.org/lkml/2021/4/29/274
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
This change allows the user to choose not to fallback to NBD mounter
when some ImageFeatures are absent with krbd driver, rather just fail
the NodeStage call.
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Currently, we recognize and warn for the provided image features based on
our prior intelligence at ceph-csi (i.e based on supportedFeatures map
and validateImageFeatures) at image/PV creation time. It might be very
much possible that the cluster is heterogeneous i.e. the PV creation and
application container might both be on different nodes with different
kernel versions (krbd driver versions).
This PR adds a mechanism to check for the supported krbd features during
mount time, if the krbd driver doesn't have the specified image feature
then it will fall back to rbd-nbd mounter.
Fixes: #478
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
this commit make use of the migration request secret parsing and set
the required fields for further nodestage operations
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
parseAndDeleteMigratedVolume() prviously clubbed the logic of
parsing of migration volume handle and then continued with the
deletion of the volume. however this commit split this
logic into two, ie parsing has been done in parseMigrationVolID()
and DeleteMigratedVolume() deletes the backend volume.
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
This commit adds a couple of helper functions to parse the migration
request secret and set it for further csi driver operations.
More details:
The intree secret has a data field called "key" which is the base64
admin secret key. The ceph CSI driver currently expect the secret to
contain data field "UserKey" for the equivalant. The CSI driver also
expect the "UserID" field which is not available in the in-tree secret
by deafult. This missing userID will be filled (if the username differ
than 'admin') in the migration secret as 'adminId' field in the
migration request, this commit adds the logic to parse this migration
secret as below:
"key" field value will be picked up from the migraion secret to "UserKey"
field.
"adminId" field value will be picked up from the migration secret to "UserID"
field
if `adminId` field is nil or not set, `UserID` field will be filled with
default value ie `admin`.The above logic get activated only when the secret
is a migration secret, otherwise skipped to the normal workflow as we have
today.
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
Thick-provisioning was introduced to make accounting of assigned space
for volumes easier. When thick-provisioned volumes are the only consumer
of the Ceph cluster, this works fine. However, it is unlikely that this
is the case. Instead, accounting of the requested (thin-provisioned)
size of volumes is much more practical as different types of volumes can
be tracked.
OpenShift already provides cluster-wide quotas, which can combine
accounting of requested volumes by grouping different StorageClasses.
In addition to the difficult practise of allowing only thick-provisioned
RBD backed volumes, the performance makes thick-provisioning
troublesome. As volumes need to be completely allocated, data needs to
be written to the volume. This can take a long time, depending on the
size of the volume. Provisioning, cloning and snapshotting becomes very
much noticeable, and because of the additional time consumption, more
prone to failures.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
for comparing the image states use the states
defined in the go-ceph avoid creating of the
deplicate const in cephcsi.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
below are the local states of the mirrored image
"unknown" -> If the image is in an error state
means data is completely synced
"error" -> If the image is in an error state
means it needs resync
"syncing"
"starting_replay"
"replaying"
"stopping_replay"
"stopped"
If the resync is successfully started which
means the image will be in "replaying" state.
we can consider "replaying" state to report
resync succesfully going on state.
we are discarding the intermediate states like
"syncing", "starting_replay" and "stopping_replay".
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
After moving moving image to trash, if `trash remove` step fails,
then external-provisioner will issue subsequent requests, in which
image will be absent in pool( will be in trash) and omap cleanup will
be done with stale image left in trash with no `trash remove` step on it.
To avoid this scenario list trash images and find corresponding id for given
image name and add a task to flatten when we encounter a ErrImageNotFound.
Fixes: #1728
Signed-off-by: Rakshith R <rar@redhat.com>
In some corner case like `re-player shutdown` the
local image will not be in error state. It would
be also worth considering `description` field to
make sure about split-brain.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
previously we were retriving clusterID using the monitors field
in the volume context at node stage code path. however it is possible to
retrieve or use clusterID directly from the volume context. This
commit also remove the getClusterIDFromMigrationVolume() function
which was used previously and its tests
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
we reuse or overload the variable name in the test execution at present.
This commit use a different variable name as initialized in each run
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
For static volume, the user will manually mounts
already existing image as a volume to the application
pods. As its a rbd Image, if the PVC is of type
fileSystem the image will be mapped, formatted
and mounted on the node,
If the user resizes the image on the ceph cluster.
User cannot not automatically resize the filesystem
created on the rbd image. Even if deletes and
recreates the kubernetes objects, the new size
will not be visible on the node.
With this changes During the NodeStageVolumeRequest
the nodeplugin will check the size of the mapped rbd
image on the node using the devicePath. and also
the rbd image size on the ceph cluster.
If the size is not matching it will do the file
system resize on the node as part of the
NodeStageVolumeRequest RPC call.
The user need to do below operation to see new size
* Resize the rbd image in ceph cluster
* Scale down all the application pods using the static
PVC.
* Make sure no application pods which are using the
static PVC is running on a node.
* Scale up all the application pods.
Validate the new size in application pod mounted
volume.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
in NodeStage operation we are flattening
the image to support mounting on the older
clients. this commits moves it to a helper
function to reduce code complexity.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
During PVC snapshot/clone both kms config and passphrase needs to copied,
while for PVC restore only passphrase needs to be copied to dest rbdvol
since destination storageclass may have another kms config.
Signed-off-by: Rakshith R <rar@redhat.com>
This commit adds the logic to detect a passed in volumeID
is a migrated volume ID and if yes, the driver connect to the
backend cluster and clean/delete the image. The logic
only applied if its a migration volume ID. The migration volume ID
carry the information like mons, pool and image name which is
good enough for the driver to identify and connect to the backend
cluster for its operations.
migration volID format:
<mig>_mons-<monsHash>_image-<imageUID>_<poolHash>
Details on the hash values:
* MonsHash: this carry a hash value (md5sum) which will be acted as the
`clusterID` for the operations in this context.
* ImageUID: this is the unique UUID generated by kubernetes for the created
volume.
* PoolHash: this is an encoded string of pool name.
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
as we are refractoring the cephfs code,
Moving all the core functions to a new folder
/pkg called core. This will make things easier
to implement. For now onwards all the core
functionalities will be added to the core
package.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
the migration nodestage request does not carry the 'clusterID' in it
and only monitors are available with the volumeContext. The volume
context flag 'migration=true' and 'static=true' flags allow us to
fill 'clusterID' from the passed in monitors to the volume Context,so
that rest of the static operations on nodestage can be proceeded as we
do treat static volumes today.
Signed-off-by: Humble Chirammal <hchiramm@redhat.com>
Currently, we delete the ceph client log file on unmap/detach.
This patch provides additional alternatives for users who would like to
persist the log files.
Strategies:
-----------
`remove`: delete log file on unmap/detach
`compress`: compress the log file to gzip on unmap/detach
`preserve`: preserve the log file in text format
Note that the default strategy will be remove on unmap, and these options
can be tweaked from the storage class
Compression size details example:
On Map: (with debug-rbd=20)
---------
$ ls -lh
-rw-r--r-- 1 root root 526K Sep 1 18:15
rbd-nbd-0001-0024-fed5480a-f00f-417a-a51d-31d8a8144c03-0000000000000003-d2e89c87-0b4d-11ec-8ea6-160f128e682d.log
On unmap:
---------
$ ls -lh
-rw-r--r-- 1 root root 33K Sep 1 18:15
rbd-nbd-0001-0024-fed5480a-f00f-417a-a51d-31d8a8144c03-0000000000000003-d2e89c87-0b4d-11ec-8ea6-160f128e682d.gz
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Log:
internal/rbd/rbd_attach.go:424:2: hugeParam: dArgs is heavy (88 bytes);
consider passing it by pointer (gocritic)
Signed-off-by: Prasanna Kumar Kalever <prasanna.kalever@redhat.com>
Currently we return a !ready status if an image
is not found when a replication resync is issued.
We also return a !ready just post issuing a resync.
The change is to ensure we return errors in these
cases for the caller to retry the operation till
we can determine we are actually resyncing, and then
return !ready with nil errors.
Part of addressing:
https://github.com/csi-addons/volume-replication-operator/issues/101
Signed-off-by: Shyamsundar Ranganathan <srangana@redhat.com>
This commit:
- modifies GetMonsAndClusterID() to take clusterID instead of options.
- moves out validation of clusterID is set or not out of GetMonsAndClusterID().
- defines ErrClusterIDNotSet new error for reusability.
- add GetClusterID() to obtain clusterID from options.
Signed-off-by: Rakshith R <rar@redhat.com>
