Couchbase Server Indexes Swelling

Couchbase Server
1

Hi again,

I have Couchbase Server Enterprise Edition 7.6.2 build 3721 and i have a serious problem.

On previous topics i mentioned few things about swelling secondary indexes, i tried many things but they keep growing (single index 100 mb) even a bucket have few mb of documents.

So there is a illogicality between index size and bucket size (x100)

I made configuration: if index fragmentation more than %10 start compaction, compaction can be 7/24

But still compaction doesn’t work, i can’t see anything on indexer.logs.

Thank you.

2

HI Kadirtemir!

How many documents do you have in the bucket? How are the indexes defined? What types of fields are indexed?

Thank you!

3

Hi!

I have 857 documents on releated bucket.

And you can see indexed fields with their size on /opt/couchbase/var/lib/couchbase/data/@2i

110M	..._idx_reportId_5186770601338626069_0.index
124M	..._idx_reportDay_1633525913979408251_0.index
110M	..._idx_archieve_id_2452013482522448870_0.index
4.0K	...#primary_17738833873087600135_0.index
110M	..._idx_reportId_reportStatus_archieveID_16538292267986105414_0.index
110M	..._idx_fiscalId_reportDay_9178269148505840638_0.index
110M	..._idx_report_status_day_device_6569965912428742003_0.index
110M	..._idx_dailyreport_archieveID_538532684115292548_0.index
110M	..._idx_reportId_status_desc_3862570136573285838_0.index
4

You can see %99 of data is trash, also you can check the indexes data size and disk size.

I did index compaction manually but still nothing changed on disk size.

image
image874×100 4.66 KB

5

Hi again,

Waiting for your support.

Thank you.

6

Hi @kadirtemir

Thanks for reaching out. Please collect and send us the logs you can do that from the UI or cbcollect_info | Couchbase Docs .

Thanks
Sai

7

Hi,

Which logs do you need?

Thank you.

8

Hi again,

waiting for your support.

9

Hi @kadirtemir

Please do a cbcollect. It will collect the necessary logs which will help us.

Thanks
Sai

10

To elaborate more we will need indexer logs in particular but some diag info that we get from cbcollect will be useful in many cases. These diag info includes golang cpu, memory profiles, state of indexes and their distribution and system information etc. Having projector and KV logs will also help sometimes if issue relates to data service. So doing a cbcollect and sending the zip file collected by it will help in debugging the issue faster.

Sorry for the delay in response.

11

Hi,

Not problem at all, i solved this issue with enabling manual compaction mode, somehow when manual compaction is disabled it doesn’t allow to run auto compaction.

After enabling manual compaction trigger it started to clean gsi indexes.

Thank you for your support.

2 Likes