Longhorn v24.09 #37
Longhorn v24.09 #37
Conversation
The minimun number of raid1 base bdevs is set to 1, so a raid1 bdev can be created with only one base bdev. Signed-off-by: Damiano Cipriani <[email protected]>
Signed-off-by: Derek Su <[email protected]>
The returned value write_string_or_name() is not really checked, so one double quote char error is not identified and the broken json string is directly sent to the client. To address the issue, the workaround is always appending one double quote char in the end of string. TODO: 1. Find the root cause 2. Check the returned value of write_string_or_name() Longhorn 6190 Signed-off-by: Derek Su <[email protected]>
Refactored the code creating raid_bdev_module_get_min_operational, distinguished intentional removal from failure in superblock base bdev state. Signed-off-by: Damiano Cipriani <[email protected]>
Implemented the functionality to add a new base bdev to an existing raid bdev, growing the size of the raid if needed. Actually only raid1 supports growing with a new base bdev. Signed-off-by: Damiano Cipriani <[email protected]>
Implemented the RPC to add a new base bdev to a raid bdev, growing the size of the raid if there isn't an empty base bdev slot. Signed-off-by: Damiano Cipriani <[email protected]>
Get a fragmap for a specific segment of a logical volume using the provided offset and size. A fragmap is a bitmap that records the allocation status of clusters. A value of "1" indicates that a cluster is allocated, whereas "0" signifies that a cluster is unallocated. Longhorn 6138 Signed-off-by: Derek Su <[email protected]>
Longhorn 6138 Signed-off-by: Derek Su <[email protected]>
Longhorn 6604 Signed-off-by: Derek Su <[email protected]>
Longhorn 6604 Signed-off-by: Derek Su <[email protected]>
Longhorn 6604 Signed-off-by: Derek Su <[email protected]>
Longhorn 6604 Signed-off-by: Derek Su <[email protected]>
Flush I/O has been implementd in raid1 simply routing the flush command to all base bdevs composing the raid bdev. Signed-off-by: Damiano Cipriani <[email protected]>
Signed-off-by: David Ko <[email protected]>
Signed-off-by: Derek Su <[email protected]>
This function creates a snapshot taking as additional parameter a list of xattrs for the newly created snapshot. Once the snapshot is created, its metadata become read only, so we can set xattrs only during the creation. Signed-off-by: Damiano Cipriani <[email protected]>
New params have been added to vbdev_lvol_create_snapshot to set xattrs in the newly created snapshot. Signed-off-by: Damiano Cipriani <[email protected]>
A new option has been added to the bdev_lvol_snapshot RPC to set xattrs in the newly created snapshot. Signed-off-by: Damiano Cipriani <[email protected]>
…a single block device
- disk-status prints the disk status by json-formatted string.
```
0000:00:03.0 (1af4 1001): Active devices: holder@vda3:dm-0,mount@vda:vda2,mount@vda:ubuntu--vg-ubuntu--lv, so not binding PCI dev
{"bdf":"0000:00:04.0","type":"virtio","driver":"virtio-pci","vendor":"1af4","numa":"unknown","device":"-","block_devices":"vdb"}
```
- disk-driver prints driver name associated with a given PCI device's BDF address.
Example:
```
bash ./setup.sh get-device-driver 0000:00:04.0
```
- bind binds a PCI device to a specific driver.
Example for binding a device:
```
export PCI_ALLOWED=0000:00:04.0; export DRIVER_OVERRIDE=uio_pci_generic; bash ./setup.sh bind
```
- unbind unbinds a PCI device from its current driver and rebind it to the original driver
Example for unbinding a device:
```
bash ./setup.sh unbind 0000:00:04.0
```
Longhorn 7672
Signed-off-by: Derek Su <[email protected]>
UNMAP I/O has been implementd in raid1 simply routing the UNMAP command to all base bdevs composing the raid bdev. Longhorn 7925 Signed-off-by: Derek Su <[email protected]>
drivers_d is already replaced with pci_bus_driver in spdk@4f8177b Signed-off-by: Derek Su <[email protected]>
A delta bitmap for fast rebuild is created and updated for a defined amout of time when the bdev of a base bdev is deleted or when it gets write I/O error. Longhorn 5573 Signed-off-by: Damiano Cipriani <[email protected]>
Older environment may not support RWF_NOWAIT, do not apply the flag if it is not supported. Change-Id: I40e83ef52aa3842c63bc41f48c4c57eaecd48ac1 Signed-off-by: Jinlong Chen <[email protected]> Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/25372 Reviewed-by: Jim Harris <[email protected]> Tested-by: SPDK CI Jenkins <[email protected]> Reviewed-by: Konrad Sztyber <[email protected]> Reviewed-by: GangCao <[email protected]> Community-CI: Community CI Samsung <[email protected]> Community-CI: Broadcom CI <[email protected]> Community-CI: Mellanox Build Bot Reviewed-by: Changpeng Liu <[email protected]>
With Linux loop device, occasionally, we continuosly get an EAGAIN error, so entering in an infinite loop. Longhorn 9700 Signed-off-by: Damiano Cipriani <[email protected]>
@DamiaSan please also remember to add a reference link to the source issue instead of the issue number only, so we can clearly understand what ongoing PRs are. |
|
@DamiaSan Is the lvstore creation stuck issue resolved? |
|
It is a new feature introduced with this commit. After this, with another commit, they disabled the functionality where it is not supported. I have simply disdabled it until we are not able to understand the problem and fix it, there shouldn't be any side effects because until Septermber the functionality was not present in code base and it is disabled in environment where it is not supported. |
|
@DamiaSan I have marked the longhorn-v24.09 as the default branch. Then, you can create PRs for updating the dependency in go-spdk-helper, longhorn-spdk-helper and longhorn-instance-manager. |
For |
Could you separate the PRs? Create separate PRs for the spdk dependency update. |
Sure that I can, but what is the benefit? In each PR there are separate commits, one for the dependency update and one for the delta bitmap. |
It is helpful when we need to check the context of the PR and its ticket. The two commits serve for different and individual purposes, so I suggest separating them. The spdk update PR should belong to longhorn/longhorn#9700. |
|
Ok, I will separate them. |
Yes, everyone's bandwidth has been very tight in recent months. Please ping other members if the PR is urgent. |
Which issue(s) this PR fixes:
Issue longhorn/longhorn#9700
What this PR does / why we need it:
New branch based on SPDK 24.09 release
Special notes for your reviewer:
Additional documentation or context