Low disk space in data service is not clearly alerting the users

Description

On a benchmarking engagement, customer allocated 2x the estimated raw data size for disk space. However their accounting didn't take into account the requirement to get 3x the total disk space. Under this condition, engine had trouble compacting the files due to lack of disk space. However there was no alert or end user indication that indicated disk space was low. We need to add more alerting to indicate disk space is low on a cluster even if the steady state without compaction is in a comfortable range.
We need a new alert that indicates that "free space on disk for vbuckets is below the total size of the bucket data on 1 or more nodes". This will clearly indicate we have an issue and confusion will be prevented.

Components

Affects versions

Fix versions

Labels

Environment

None

Release Notes Description

None

Activity

Raju Suravarjjala July 8, 2020 at 9:18 PM

Bulk closing invalid, duplicate and won't fix bugs

Shivani Gupta June 26, 2018 at 1:31 PM

Agree . I am going to close this MB.

Sarath Lakshman June 26, 2018 at 4:18 AM

Our target is to maintain less than 50% fragmentation (< 1.5x space amplification) and incremental compaction.

Shivani Gupta June 25, 2018 at 7:58 PM

We are going to have a new storage engine(called Magma) for DGM/High Data Density. Compaction overhead will be less with Magma.

 how much extra space will be required for compaction with Magma? Or is compaction always incremental and hence very little space overhead?

Don Pinto [X] June 22, 2018 at 5:20 AM
Edited

, this problem might become worse with DGM/High data density. assigning to you to triage further or close if no longer relevant

Cannot Reproduce
Pinned fields
Click on the next to a field label to start pinning.

Details

Assignee

Reporter

Priority

Instabug

Open Instabug

PagerDuty

Sentry

Zendesk Support

Created January 19, 2016 at 4:26 PM
Updated July 8, 2020 at 9:18 PM
Resolved June 26, 2018 at 1:34 PM
Instabug