Cleaner FAQ (4044794)

Cleaner FAQ

• The Cleaner is not working

• Space is not being reclaimed

• Not enough space is being reclaimed

• The performance is slower than expected

• The system uses more resources than expected

The DR is getting filled, or there are a lot of jobs that fail, or possibly less space than expected is being reported, or observed backup job performance is slower than expected

Progress

The cleaner has many functions. We recommend to read the Cleaner best practices on the support website for an insight on how to configure the Cleaner. https://support.quest.com/dr-series/dr4100/technical-documents

One of the most important aspects to running the Cleaner is time. A short description is that during a backup-job the data is sent to the DR, it is then queued for handling, like a Inbox. Then the Cleaner starts and looks over the data for duplicated information. What would happen after that is that the DR needs to compress the data that is stored. After compression the Cleaner starts to reclaim data before finally encrypting the volumes (if enabled). This can be viewed by using the CLI command:

stats --cleaner

 

The important number here is at the end of the output, "Estimated logical bytes left". This would indicate how large the workload is for the Cleaner at any given moment. The larger the number, the more time the system needs to reclaim space. Every job, replication, backup or reclamation, adds to the list of logical bytes. To check if the Cleaner is working, this number should always be decreasing, or be as close to 0 as possible. Please bear in mind that when savesets expire, the space is not reclaimed at first, reclamation is one of the last phases of the Cleaner tasks. Always wait until there is 0 bytes left to clean before evaluating the effectiveness.

If the estimated bytes left increases, this means that the Cleaner needs more time to handle the incoming jobs. There is a balance that needs to be found between running backups, then cleaning the data, then replicating data (if applicable). The larger the backup job (amount of data, or amount of clients) the more time needed for cleaning. It is recommended to not run any jobs while cleaning.

Reclamation

The Cleaner will not be able to reclaim all space that has been marked for deletion. To know how much data needs to be expired, first calculate the savings ratio. In the CLI, use this command:

stats --system

At the end of the output "Total savings" is visible and this represents the duplication ratio. The space can be calculated like this: "Reclaimed space" approx "Expired data" * ((100 - "Total savings") / 100).

 

Following that example, calculating 1Tb data to be expired with a savings of 90% equalling approximately 100Gb of potential savings. This calculation is heavily depending on how much unique data has been stored. For example, if there is a full backup of a system that has little to no changes, it may be noticed that not much space is reclaimed at all. This would be normal.

Perfomance impact

The Cleaner can be run in three different modes:

1. Idle (little impact to backup performance)
2. Scheduled (medium impact on backup performance)
3. Forced (big impact on backup performance)

When the DR does not have any IO for a period of time, the Cleaner will run automatically. This is "Idle" mode. Whenever the IO restarts, the Cleaner will stop running and all system resources are focused on the incoming jobs. This will increase the "Estimated bytes left" count. If there is a schedule (according to best practices), the Cleaner will be a bit more aggressive and will run during the scheduled time, and when there is IO the Cleaner will run at a lower priority. This might increase the "Estimated bytes left" count and will have a smaller performance impact. Running the Cleaner manually, either by using the CLI, or by using the webGUI will force the Cleaner to run, and it will run until it is finished. This might increase the "Estimated bytes left" but should decrease the count, but it will have a huge impact on performance. It is recommended to stop any and all jobs to the system while running the Cleaner manually. Running the Cleaner manually should only be used when there is a backlog or the Cleaner has not completed a run in more than a week.