Manual/VirtOn PBS (For the sake of mystery.)/6. PBS Frun & Garbage collection (Prune & Garbage Collection)
Home
VirtOn PBS DocumentationVersion 1.0User Guide

6. PBS Frun & Garbage collection (Prune & Garbage Collection)

On this page

6. PBS Frun & Garbage collection (Prune & Garbage Collection)

  • Frun, please. (Prune)
    • Definition: What is it? Set a 'survival policy'(Retention Policy)'depending on whether the storage period has expired or is unnecessary. Back up snapshot.(The index) The task of disconnecting and deleting from the listThat's it.
  • The collection of garbage (Garbage Collection)
    • Definition: What is it? Because of the work of Frun, you can no longer see any of the backup snapshots.(All the name tags are gone.) The actual data slice.(Chunk) Find the ears on the disk and physically delete them.That's it.

6.1 Work order(Prune) Policy Function

  • Purpose
    • PBS Backup policy(Retention Policy) It's a feature to register/correct/delete and run immediately when needed.
  • The core operation
    • Policy storage: PBS Create/renew the prune job settings
    • Policy list view: Save prune policy table view
    • Policy selection: Load the selected policy into the form(Fixed mode)
    • Delete policy: Delete policy of selection
    • Manual run: Prune run immediately according to the selected policy
    • Status check: Recently organized view/state/next execution etc.

6.1.1 Explanation of each input item

  • PBS The server.
    • The target. PBS Select the connection server
    • The list of servers is registered PBS Load based on connection settings
  • The data store.
    • Prune object datastore
    • The default datastore of the selected server is reflected as the default value
  • The namespace.
    • Prune applies range namespace specified
    • If you empty it, Root.(- (Root)) The standard.
  • Maximum depth. (max-depth)
    • Under namespace browsing depth
    • If 0 is the current namespace function
  • Calendar is fixed. (schedule)
    • The cron/preset schedule
    • Yeah, hourly, daily, two and a half, mon..fri \*:00
  • Conservation of the water(Keep policy)
    • Recently kept (keep-last)
    • Keep it on the hour. (keep-hourly)
    • Keep it up to date. (keep-daily)
    • Keep it for a week. (keep-weekly)
    • Keep it monthly. (keep-monthly)
    • Keep it yearly. (keep-yearly)
    • Leave a snapshot of each condition and the rest of the prune is subjected to
  • Explanation (comment)
    • Operating notes/policy description text
  • Activated (enabled)
    • Check time schedule is active.
    • The policy of disengagement remains and the implementation is inactive.
  • I.D. of work (id)
    • Policy-specific identifiers
    • When you empty it, it automatically generates
    • In the modification mode, the fixed(Preventing the change of identifier)

5.1.2 Mode of operation

  • The generated mode
    • Save the new policy
    • The job ID can be entered directly(Or automatically.)
  • Fixed mode
    • Save changes after loading the existing policy value
    • Work ID is read-only.
    • Reverse the modification to the generated mode

6.1.3 Run/run point

  • Run the Prune
    • Logical solution(Remove the reference) Center
    • The disk capacity can be limited immediately.
  • Relationship with GC
    • The actual physical capacity recall is carried out at the GC stage
    • Usually Prune -> GC order operation recommended
  • Beware of it.
    • If the policy values are set aggressively, the recovery points can accelerate, so the operating parameters are(Months and weeks and months and years) It's safe to set and apply first.

6.2 Coverage collection(Garbage Collection) And the manual execution function

  • Purpose
    • Set/delete the GC schedule, and arrange work when needed(Prune) Or a collection of covers.(GC) I'm going to run it manually right away.
  • The core operation
    • GC schedule storage: GC schedule registration/correction by datastore
    • Running by hand:
      • Work is organized and executed(Prune run immediately)
      • Run the coverage collection(GC run immediately)
    • Running tracking: Upid-based status/rate of progress/results shown in real time(SSE)

6.2.1 Input/display item by item description

  • Select the server
    • Current subject PBS Server display(For reading.)
  • The data store.
    • GC/Activated Running Object Data Store Display(For reading.)
  • GC schedule
    • I'm not going to lie.(No, hourly, daily, 02:30 etc.) Or directly enter the option
    • Direct input when selecting the Cron/Pryset string directly input
  • GC constant save button
    • The current GC constant is PBS Save to
    • No, when you select the schedule, remove it.(Inactivity)

6.2.2 Manual run button

  • Work is organized and executed
    • Save the prune policy(The job ID) To execute immediately
    • If you don't have a policy option/ID, execute blocked.
  • Run the coverage collection
    • GC immediately run to the current datastore
    • Re-evaluate the GC status section after execution

6.2.3 Running state card(Recent work)

  • Recent work
    • PRUNE or GC
  • UPID
    • PBS The task-specific identifier
    • The default value for status monitoring/stream tracking
  • It's in the air.
    • idle, running, stopped, etc.
  • The rate of progress
    • Logging/SSE based percentage display
  • The results.
    • OK or fail code(exitstatus)
  • The progress bar.
    • 0 to 100% visualization

6.2.4 Movement flow

  1. Target verification by server/data store
  2. Save the GC schedule when needed
  3. Running manually(PRUNE or GC)
  4. Subscribe to SSE stream after receiving UPID
  5. The progress rate, the status, the results, the update.
  6. Completed time toast + GC state reactivation

6.2.5 Operating details

  • Prune ≠ instantaneous capacity recovery
    • Prune is a logical solution, reflecting the real-space recruitment at the GC stage.
  • Less than 24 hours, Zinc.
    • GC repeat run may not be immediately recovered.
  • Load management
    • We recommend separating the backup/activation time and the GC time.
    • In environments with high I/O competition, it supports the manual GC footprint.

6.3 List of policies for the collection of coverage and the organization of stored work

6.3.1 Coverage collection status (GC Status)

  • Purpose
    • I'll check the GC execution results of the current data store and the next execution information.
  • Display items
    • Schedule: GC schedule(cron/preset)
    • Recent GC: Last GC runtime view
    • Duration: Last GC work time
    • Last run:(OK, ERROR and so on.)
    • Next run: next booking run time
    • Data removed: The actual volume recovered from the last GC
    • Data in reserve: The potential data capacity that has not been recovered yet
  • How it works.
    • It's a server/data store selection.
    • After GC is executed, the status table is updated to the latest value.
  • Operating point
    • It's not uncommon for data to be deleted even if it's zero.(Effects of Conservation/Heritage Conditions).
    • If the data in reserve continues to increase, we need to check the schedule/reservation policy.

6.3.2 List of policies stored (Saved Prune Policy List)

  • Purpose
    • Registered work order policy(Prune Job) It's a control table for searching, filtering, selecting, correcting, deleting.
  • The core function
    • Searching: Work ID/data store/name space by speed of search
    • I'm going to get you.: Schedule/State Standards List Reduction
    • Select the line: Load the selected policy into the form and enter the modification mode
    • Select and delete: Delete the selected policy(Based on verification dialog)
  • Column to display
    • Work ID, data store, namespace, maximum depth
    • Calendar, recent/hourly/daily/weekly/monthly/annual conservation
    • Recent order, duration, state, next run, comment
  • How it works.
    • The list data is PBS It's rendered based on the results of the policy review.
    • The line selection form is Fixed modeWhen you switch to, cancel, The generated modeSo we're going back to the ro.
  • Operating point
    • After the policy is saved, corrected, deleted, the list is immediately re-edited and matched to the screen-reality status.
    • Root namespace is a UI that's a(Root) It's a label that keeps it identifiable.
  1. Check whether the GC status is current and the recent GC peak
  2. Review the schedules/preservation values of the policy subject in the policy list of storage
  3. When necessary, modify the policy and then execute it manually.(Prune → GC)
  4. After execution, the data is removed, the data in backup, check to the workforce log