-
Notifications
You must be signed in to change notification settings - Fork 542
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Separating Ext2/3/4 from filesys and using lsblk -rpo instead /proc/mounts #3502
base: main
Are you sure you want to change the base?
Conversation
Congratulations! One of the builds has completed. 🍾 You can install the built RPMs by following these steps:
Please note that the RPMs should be used only in a testing environment. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
With this refactor, we are now collecting potentially the same set of items from 2 separate plugins, and we are typically not keen on that.
We'll also need to double check to see if any automation that our downstream have that interrogates files collected in filesys
in order for them to be removed from there.
I am sure other core reviewers may have other things on this
I'm not sure if this should go in the same commit or a separated one. My feeling is that it should be a second commit, but lets hear other's opinions. We are still using /proc/mounts in filesys.py:
So it may be worth removing that code completely from filesys now, since you'll be getting it via this new plugin. |
Yeah sorry about that forgot to push filesys where the Ext2/3/4 are removed. I will update the pull request. |
Same on our side, but I see a lot of benefits on these changes, specially separating ext from filesys like xfs is -that is, with my storage/filesystem engineer hat, I can find it really really useful. We need to find a way to make sure these changes are properly propagated to users using automation. Some kind of policy for coexistence and maybe duplication for some time, in terms of minor releases. Other options, but I imagine these are less agreeable, are:
I don't think we've ever done these kind of things in sos, I guess to avoid interlinking plugins. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
some other changes that came into mind, based on contribution guidelines
It can happen the FS of interest is not actually mounted and then we have no info about it. This is revorks of how we should get list of devices. It adds a new __init__.py fuction to decrease code duplication get_dev_by_fstype(self,fstype). It is using lsblk instead of /proc/mounts. Last change is separation of Ext2/3/4 info into separate module from filesys. Signed-off-by: Lukas Herbolt <[email protected]>
I hope I add all the changes. About the duplicate info, we can append to the filesys dumpe2fs some note about being deprecated in next X.Y releases. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Mostly a couple minor points below, however I think it warrants fleshing out moving the discovery of filesystems to SoSReport
, as part of our _get_hardware_devices()
bits, so that we generate this information once, and have plugins (and potentially any other part of sos) reference that collection.
def get_devices_by_fstype(self, fstype): | ||
"""Get list of devices based on fstype and returns empty array in case | ||
of failure. | ||
""" | ||
dev = [] | ||
all_devs = self.exec_cmd("lsblk -rpo NAME,FSTYPE") | ||
if (all_devs['status'] == 0 and not all_devs['truncated']): | ||
if all_devs['status'] == 0: | ||
for line in all_devs['output'].splitlines(): | ||
if (fstype in line) and (line.split()[0] not in dev): | ||
dev.append(line.split()[0]) | ||
return dev |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm torn about having this move from Plugin
to SoSReport
, and have it run alongside our device introspection, like for block devs. Granted, these aren't hardware devices, but it's in the same vein as what we do for storage and network interfaces.
That way we run this once, when sos first initializes, and then in Plugin
this function would instead simply filter through an already-made list/dict/whatever for those matching the requested fstype
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That's a good point. I will into it, the main idea here is to reuse the info, not just for FS but also md.py can benefit from it as the FSTYPE in limited to FS as per blkid -k
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm torn about having this move from
Plugin
toSoSReport
, and have it run alongside our device introspection, like for block devs. Granted, these aren't hardware devices, but it's in the same vein as what we do for storage and network interfaces.That way we run this once, when sos first initializes, and then in
Plugin
this function would instead simply filter through an already-made list/dict/whatever for those matching the requestedfstype
.
So I have the work almost ready to have this merged into _get_hardware_devices()
, but there is thing. If I create nested dict like:
{'storage': {'block': ['data0', 'data1'], 'fstype': {'unknown': ['block0','block1'],
'vfat': ['block2']}}}
The function recursive_dict_values_by_key()
will not correctly return the device list, only if I put it into same level as storage and network dict are. So the function does not seems to recursive properly. Any pointers what could be there wrong?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sorry I should have been clearer - my intent was to suggest that this be at the same level as storage
, network
, or similar. So plugins would access it via something like for fs in self.devices['filesystems']['xfs']: ...blah...
.
Again, this is a bit of a misnomer placing it under devices
, so if you prefer we can also make a separate attribute to hold these, so that plugins could do something like for fs in self.filesystems['xfs']:
if allfs: | ||
for fs in allfs: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The conditional is superfluous, if allfs
is an empty list, python just continues on, skipping whatever is in the for-loop.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Are you suggesting something like:
def setup(self):
allfs = self.get_devices_by_fstype('xfs')
for fs in allfs:
self.add_cmd_output(f"xfs_info {fs}", tags="xfs_info")
self.add_cmd_output(f"xfs_admin -l -u {fs}")
if not allfs:
mounts = '/proc/mounts'
ext_fs_regex = r"^(/dev/.+).+xfs\s+"
for dev in zip(self.do_regex_find_all(ext_fs_regex, mounts)):
for e in dev:
parts = e.split()
self.add_cmd_output(f"xfs_info {parts[1]}",
tags="xfs_info")
self.add_cmd_output(f"xfs_admin -l -u {parts[0]}")
self.add_copy_spec(self.files)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yep, that's correct. We could even button it up a little more by for fs in self.get_devices_by_fstype('xfs'):
provided there is nothing else using allfs
later on.
if allfs: | ||
for fs in allfs: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Same here, we can drop the conditional and unnest the entire block by one level here.
if allfs: | ||
for fs in allfs: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ditto here.
Automation should be using This isn't really that different from other plugins we've broken up in the past, like |
option_list = [ | ||
PluginOpt('dumpe2fs', default=False, desc='dump filesystem info'), | ||
PluginOpt('frag', default=False, | ||
desc='collect filesystem fragmentation status') | ||
] | ||
|
||
def setup(self): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, missed this on my first review. This seems like a copy/paste error. The options_list shouldn't go within setup()
, and we've got a duplicate definition of the setup()
method, which would break several things. If I'm reading everything right on the github webui, I think you just need to remove these lines entirely.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah copy&paste error
I will finalize this one after ACK/NACK/CHANGE of #3538 |
[xfs,filesys,ext] Get list of devices by lsblk -lpo NAME,FSTYPE.
It can happen the FS of interest is not actually mounted and then we have no info about it. This is revorks of how we should get list of devices. It adds a new init.py fuction to decrease code duplication get_dev_by_fstype(self,fstype). It is using lsblk instead of /proc/mounts. Last change is separation of Ext2/3/4 info into separate module from filesys.
Please place an 'X' inside each '[]' to confirm you adhere to our Contributor Guidelines