mbox series

[nfs-utils,0/8] mountstats/nfsiostat: bugfixes for iostat

Message ID 20250130142008.3600334-1-sorenson@redhat.com (mailing list archive)
Headers show
Series mountstats/nfsiostat: bugfixes for iostat | expand

Message

Frank Sorenson Jan. 30, 2025, 2:19 p.m. UTC
This patchset is intended to fix several bugs in nfsiostat and
'mountstats iostat', in particular when an <interval> is specified.

Specifically, the patches address the following issues when printing
multiple times:

* both nfsiostat and 'mountstats iostat':
   - if <count> is also specified, the scripts sleep an additional
     <interval> after printing <count> times, and parse mountstats
     unnecessarily before checking the <count>

   - if no nfs mounts are present when printing, the scripts output
     a message that there are no nfs mounts, and exit immediately.

     However, if multiple intervals are indicated, it makes more sense
     to output the message and sleep until the next iteration; there
     may be nfs mounts the next time through.

   - if mountpoints are specified on the command-line, but are not
     present during an iteration, the scripts crash.


 * 'mountstats iostat':
   - if an nfs filesystem is unmounted between iterations, the
     script will crash attempting to reference the non-existent
     mountpoint in the new mountstats file (or if another filesystem
     type is mounted at that location, will instead crash while
     comparing old and new stats)

   - new nfs mounts are not detected


 * nfsiostat:
   - if a new nfs filesystem is mounted between iterations, the
     script will crash attempting to reference the non-existent
     mountpoint in the old mountstats file

   - if a mountpoint is specified on the command-line, but topmost
     mount there is not nfs, the script crashes while trying to
     compare old and new stats


To address these issues, the patches do the following:
 * when printing diff stats from previous iteration:
   - verify that a device exists in the old mountstats before referencing it
   - verify that both old and new fstypes are the same

 * when filtering the current mountstats file:
   - verify the device exists in the mountstats file before referencing it
   - filter the list of nfs mountpoints each iostat iteration

 * check for empty device list and output the 'No NFS mount points found'
    message, but don't immediately exit the script

 * merge the infinite loop and counted loop, and (for the counted loop)
    decrement and check the count before sleeping and parsing the mountstats
    file


Frank Sorenson (8):
  mountstats/nfsiostat: add a function to return the fstype
  mountstats: when printing iostats, verify that old and new types are
    the same
  nfsiostat: mirror how mountstats iostat prints the stats
  nfsiostat: fix crash when filtering mountstats after unmount
  nfsiostat: make comment explain mount/unmount more broadly
  mountstats: filter for nfs mounts in a function, each iostat iteration
  mountstats/nfsiostat: Move the checks for empty mountpoint list into
    the print function
  mountstats/nfsiostat: merge and rework the infinite and counted loops

 tools/mountstats/mountstats.py | 102 ++++++++++++++++--------------
 tools/nfs-iostat/nfs-iostat.py | 110 +++++++++++++--------------------
 2 files changed, 100 insertions(+), 112 deletions(-)

Comments

Chuck Lever Jan. 30, 2025, 2:28 p.m. UTC | #1
On 1/30/25 9:19 AM, Frank Sorenson wrote:
> This patchset is intended to fix several bugs in nfsiostat and
> 'mountstats iostat', in particular when an <interval> is specified.
> 
> Specifically, the patches address the following issues when printing
> multiple times:
> 
> * both nfsiostat and 'mountstats iostat':
>    - if <count> is also specified, the scripts sleep an additional
>      <interval> after printing <count> times, and parse mountstats
>      unnecessarily before checking the <count>
> 
>    - if no nfs mounts are present when printing, the scripts output
>      a message that there are no nfs mounts, and exit immediately.
> 
>      However, if multiple intervals are indicated, it makes more sense
>      to output the message and sleep until the next iteration; there
>      may be nfs mounts the next time through.
> 
>    - if mountpoints are specified on the command-line, but are not
>      present during an iteration, the scripts crash.
> 
> 
>  * 'mountstats iostat':
>    - if an nfs filesystem is unmounted between iterations, the
>      script will crash attempting to reference the non-existent
>      mountpoint in the new mountstats file (or if another filesystem
>      type is mounted at that location, will instead crash while
>      comparing old and new stats)
> 
>    - new nfs mounts are not detected
> 
> 
>  * nfsiostat:
>    - if a new nfs filesystem is mounted between iterations, the
>      script will crash attempting to reference the non-existent
>      mountpoint in the old mountstats file
> 
>    - if a mountpoint is specified on the command-line, but topmost
>      mount there is not nfs, the script crashes while trying to
>      compare old and new stats
> 
> 
> To address these issues, the patches do the following:
>  * when printing diff stats from previous iteration:
>    - verify that a device exists in the old mountstats before referencing it
>    - verify that both old and new fstypes are the same
> 
>  * when filtering the current mountstats file:
>    - verify the device exists in the mountstats file before referencing it
>    - filter the list of nfs mountpoints each iostat iteration
> 
>  * check for empty device list and output the 'No NFS mount points found'
>     message, but don't immediately exit the script
> 
>  * merge the infinite loop and counted loop, and (for the counted loop)
>     decrement and check the count before sleeping and parsing the mountstats
>     file
> 
> 
> Frank Sorenson (8):
>   mountstats/nfsiostat: add a function to return the fstype
>   mountstats: when printing iostats, verify that old and new types are
>     the same
>   nfsiostat: mirror how mountstats iostat prints the stats
>   nfsiostat: fix crash when filtering mountstats after unmount
>   nfsiostat: make comment explain mount/unmount more broadly
>   mountstats: filter for nfs mounts in a function, each iostat iteration
>   mountstats/nfsiostat: Move the checks for empty mountpoint list into
>     the print function
>   mountstats/nfsiostat: merge and rework the infinite and counted loops
> 
>  tools/mountstats/mountstats.py | 102 ++++++++++++++++--------------
>  tools/nfs-iostat/nfs-iostat.py | 110 +++++++++++++--------------------
>  2 files changed, 100 insertions(+), 112 deletions(-)
> 

Hi Frank, I browsed this series briefly. I'm not sure that counts as a
full review, but it all seems sensible to me. Thanks for the clear
explanations!