Set de-fragmentation schedules

De-fragmentation is a required process as fragmented data can lead to performance issues. Users can schedule to run this process during periods of low workloads, avoiding any interference and improving performance.

Limitations:

• Not supported with the PX-StoreV2 backend

Set a defragmentation schedule

Users can use CLI commands to set up periodic defragmentation schedules during periods of low workloads. Multiple defragmentation schedules can be configured within a cluster; however, only one schedule is allowed per node. The --include-nodes option allows you to specify a list of nodes that the defrag schedule will run on.

Use the following command to set up a defragmentation schedule on a node:

pxctl cluster defrag schedule create \
--start-time <start-time> \
--max-duration-minutes <duration-of-defrag> \
--include-nodes <nodes-to-be-included-in-the-defrag-run>
--node-selector <label1=value1,label2=value2>

important

Portworx inteprets the start-time in UTC, not in the system's local time zone. Therefore, when configuring a defragmentation schedule, ensure that you convert your intended time to UTC when defining the schedule.

The following flags can be used to set up the schedule:

• --start-time: Specifies the start time of the run and can be set in the following formats:
  • daily=19:15 (starts at 19:15 UTC every day)
  • weekly=Sunday@19:15 (starts at 19:15 UTC every Sunday)
  • monthly=22@19:15 (starts at 19:15 UTC on the 22nd day of every month)
• --max-duration-minutes: Specifies the duration of each run (end time = start time + max duration).
• --include-nodes and --exclude-nodes: Specify the comma-separated UUIDs of the nodes to be included or excluded from the run (optional).
• --max-nodes-in-parallel: Specifies how many nodes will run the schedule concurrently. The default value is 1. For example, if set to 3, the schedule will run on 3 nodes in parallel (optional).
• --node-selector : Specifies the nodes with listed labels on which schedule will run on.

Configure your defrag schedules cluster-wide

You can configure all your defragmentation schedule in your cluster with two options:

• Chunk size: The step size of each defragmentation command. The default value is 32 MB.
• Pool full threshold: A percentage threshold for pool space usage. If any pool exceeds this threshold, defragmentation will be disabled for that pool. The default value is 70%.

To configure these options, use the following commands:

pxctl cluster options update --defrag-schedule-chunk-size <chunk-size>

Successfully updated cluster-wide options

pxctl cluster options update --defrag-schedule-pool-full-threshold <threshold-percentage>

Successfully updated cluster-wide options

To view the current configurations, run the following command:

pxctl cluster options list | grep Defrag

Defrag schedule chunk size (MB)             : 64
Defrag schedule pool full threshold (%)     : 75

Check the status of a defragmentation schedule

Once the schedule is set, you can use the following command to list your defragmentation schedules:

pxctl cluster defrag schedule show

Defragmentation Schedules:
Schedule ID         : 611122756865623299
Create Time         : 2024-03-08 23:49:12
Frequency           : daily
Start Time          : 23:50
End Time            : 23:51
Nodes Included      : dbca66e4-xxx-xxx, dea00c74-xxx-xxxx
One Iteration Only  : False
Max Number of Nodes Running in Parallel: 2

To check the status of the defragmentation operation for each pool, run:

pxctl cluster defrag status

Running Schedule : 611122756865623299 (daily=23:50, 1 minute)
NODE ID          NODE IP         SCHEDULER NODE NAME POOL ID           ITERATIONS POOL USED SPACE LAST START TIME     LAST COMPLETE TIME   RUNNING  PROGRESS
dbca66e4-xxx-xxx 10.13.185.207   N/A                 ce2cec09-5873-xxx 5           4%             2024-03-08 23:50:13  2024-03-08 05:12:54  Yes       0%

Delete a defragmentation schedule

You can delete schedules using the following methods:

• Use the --one-iteration-only flag while creating a schedule to automatically delete it after one full iteration:

  pxctl cluster defrag schedule create \
  --start-time <start-time> \
  --max-duration-minutes <max-duration-of-the-run> \
  --max-nodes-in-parallel <number-of-nodes> \
  --include-nodes <uuid-1>,<uuid-2> \
  --one-iteration-only

• Delete a single schedule:

  pxctl cluster defrag schedule delete <schedule-id>

• Delete all schedules:

  pxctl cluster defrag schedule clean-up