Splunk SOAR

What is the proper way to purge SOAR containers/events?

ouss
Explorer

Hello,

What is the proper way to purge Splunk SOAR/phantom containers from the database. It seems that deleting a container only hides it from the UI. 

Is there a way  to purge containers with certain filters for example purge containers where label="secific_label" and created before 6 months ... ?

 

 

Labels (1)
Tags (1)
0 Karma
1 Solution

phanTom
SplunkTrust
SplunkTrust

@ouss took me a while to find the docs but here: https://my.phantom.us/kb/80/ 

optional arguments:
-h, --help show this help message and exit
-b, --list-labels Lists available container labels and exits
-d, --dry-run Do not actually delete, just show output
--non-interactive Do not block on user input. This flag is suitable for running as part of an unsupervised script.
-c CHUNK_SIZE, --chunk-size CHUNK_SIZE
Max number of containers to delete in a single transaction. Max 10000
-r MAX_RETRY_COUNT, --max-retry-count MAX_RETRY_COUNT
Max number of retries in the case of error.

Filters:
Use these to select the containers you want to delete. Supply at least one.

-i IDS, --ids IDS Comma separated list of IDs of the container to delete
-l LABEL, --label LABEL
Only delete containers with this label
-m MATCHING, --matching MATCHING
Title matches this string (case insensitive)
--before BEFORE Only delete containers last updated before this date/time.
--after AFTER Only delete containers last updated after this date/time.
--status STATUS Only delete containers having these status values (comma separated list).

Not sure why it's not on Splunk docs yet but I have raised a feedback request to get it ported across. 

Please mark as solution if this helps. 

Tom

View solution in original post

depuy_ohsu
Observer

Googled to try and find a solution to this problem and came across this answer. Also found this Splunk docs article: https://docs.splunk.com/Documentation/SOARonprem/5.5.0/Admin/DataRetention

 

Would the above accomplish not wanting containers and their data past X years? Would be really great if this was the case.

0 Karma

Greenapps
Engager

Hey all,

Im not a postgre or linux master but I found something. 
Based on this article:

https://docs.splunk.com/Documentation/Phantom/4.10.7/Install/UpgradeOffline

I notice how to run vacuumdb in phantom Centos 7 environment.
In the article there is a line, I know that this arcitle is for old version 4 but give me some ideas how to run vacuumdb.

To improve database performance, after completing the upgrade, run: su - postgres -c '/usr/pgsql-11/bin/vacuumdb -h /tmp --all --analyze-in-stages'

After that there is the similar command but from phenv.

/<PHANTOM_HOME>/bin/phenv /<PHANTOM_HOME>/usr/postgresql/bin/vacuumdb -h /tmp --all --analyze-in-stages

Based on this i understand that --analyze-in-stages - only show statistics.

After that I open folder:

cd  /opt/phantom/usr/postgresql/bin/

Next:

phenv vacuumdb -h /tmp --all --analyze-in-stages
But this command shows me only stats.

But from this article I notice that vacuumdb command has parameter -full
https://stackoverflow.com/questions/51204561/postgres-vacuum-doesnt-free-up-space

Lets check vacuumdb documentation:
https://www.postgresql.org/docs/current/app-vacuumdb.html

--analyze-in-stages

Only calculate statistics for use by the optimizer (no vacuum), like --analyze-only. Run three stages of analyze; the first stage uses the lowest possible statistics target (see default_statistics_target) to produce usable statistics faster, and subsequent stages build the full statistics.

This option is only useful to analyze a database that currently has no statistics or has wholly incorrect ones, such as if it is newly populated from a restored dump or by pg_upgrade. Be aware that running with this option in a database with existing statistics may cause the query optimizer choices to become transiently worse due to the low statistics targets of the early stages.

--all

Vacuum all databases.

--full

Perform “full” vacuuming.

 

 

 


So for general what was done to remove containers and reduce db disc space in Splunk SOAR 5.2:

1. cd /opt/phantom/bin/
2. phenv python /opt/phantom/bin/delete_containers.pyc --label test --before "2020-01-01T12:00:00Z"

3. cd /opt/phantom/usr/postgresql/bin/

4. phenv vacuumdb -h /tmp --all --full

This reduce a space in my case.

! Warning!,

Im not sure if this method is proper, but it works in my case. You can test it on your test environment  and confirm if it works for you also. Maybe someone from Splunk Suport will be able to confirm this solution.

 General summary:
phenv vacuumdb -h /tmp --all

This command - "The standard form of VACUUM removes dead row versions in tables and indexes and marks the space available for future reuse. However, it will not return the space to the operating system,"


But this command:
 phenv vacuumdb -h /tmp --all --full

Although VACUUM FULL can be used to shrink a table back to its minimum size and return the disk space to the operating system, there is not much point in this if the table will just grow again in the future. Thus, moderately-frequent standard VACUUM runs are a better approach than infrequent VACUUM FULL runs for maintaining heavily-updated tables.

I hope it helps.

dbkuhn
Explorer

Hi All, 

Would like to re-open this discussion as what OP is seeing with disk space is a constant pain point for me too. $PHANTOM_HOME starts creeping up periodically eventually getting up to >95%. Regardless of what I am removing with the aforementioned scripts (I have semi automated and run manually from time to time). It does not help...only solution I have found is to continually grow the FS.  I have come to the conclusion that a lot of this space is eaten up temporary files created by DB queries. 

./data/db/base/pgsql_tmp/pgsql_tmp2178.xxxxx

 

Currently have over 8200 of these files in this directory. lol

 

All an assumption though. 

 

Maybe I need to understand Postgres more. Just curious if anyone has additional input on this too? 


Thanks,

 

Doug

 

 

 

 

0 Karma

phanTom
SplunkTrust
SplunkTrust

@ouss great question, honestly I am not sure as not too aware of postgres workings tbh. 

I would also say I raised the request and the docs have been moved to here now: https://docs.splunk.com/Documentation/SOARonprem/5.3.1/Admin/DeleteContainers

phanTom
SplunkTrust
SplunkTrust

@ouss took me a while to find the docs but here: https://my.phantom.us/kb/80/ 

optional arguments:
-h, --help show this help message and exit
-b, --list-labels Lists available container labels and exits
-d, --dry-run Do not actually delete, just show output
--non-interactive Do not block on user input. This flag is suitable for running as part of an unsupervised script.
-c CHUNK_SIZE, --chunk-size CHUNK_SIZE
Max number of containers to delete in a single transaction. Max 10000
-r MAX_RETRY_COUNT, --max-retry-count MAX_RETRY_COUNT
Max number of retries in the case of error.

Filters:
Use these to select the containers you want to delete. Supply at least one.

-i IDS, --ids IDS Comma separated list of IDs of the container to delete
-l LABEL, --label LABEL
Only delete containers with this label
-m MATCHING, --matching MATCHING
Title matches this string (case insensitive)
--before BEFORE Only delete containers last updated before this date/time.
--after AFTER Only delete containers last updated after this date/time.
--status STATUS Only delete containers having these status values (comma separated list).

Not sure why it's not on Splunk docs yet but I have raised a feedback request to get it ported across. 

Please mark as solution if this helps. 

Tom

ouss
Explorer

Hello @phanTom,

you are right the script works well and the containers data is purged from the database. However, I am noticing that the folder "/phantom/data/db/pg_wal" (45G) is huge compared to "/phantom/data/db/base" (4.5 G). From searchihng in Postgres forums it seems that pg_wal is supposed to be cleaned automatically by postgres ... any ideas on why this is not happening or from where I coud start troubleshooting 🙂 thank you in advance.

 

  

 

0 Karma

ouss
Explorer

Thank you for your reply @phanTom , I have executed this script and deleted approximately 10k containers with the artifacts, playbook runs and action runs associated to them ... However I do not seem to notice any changes on disk Space usage, (by the way my goal of purging containers is to free up some disk space).  Is the script deleting the data from the database or should I follow further steps to delete this data from the database?

 

Tags (1)
0 Karma

dbkuhn
Explorer

I had a case open with support on this same issue as continually adding space to the file system was not viable.  Support indicated that cleaning out containers and all related 'stuff' was fine and all...but there is a need to reclaim space from the Postgres DB after running the cleanup commands for containers and related items (artifacts, etc). I also learned that I had A TON of Indicators and indicator_artifact_records eating up space in the DB as well that needed to be removed.

Once competed, we utilized the the Postgres VACUUM function.

https://www.postgresql.org/docs/current/sql-vacuum.html

 

There is a way to see top offenders of bloat in the database too, which will help show what's eating up space.  Required a ~140 line query to add to the command (not sure if that should be shared here ). This is an example of what i ran to get that data: 

cat /tmp/phantom/psup_check_bloat.query | $PHANTOM_HOME/phantom/bin/phenv python $PHANTOM_HOME/phantom/www/manage.py dbshell  > /tmp/pg_bloat.out
 
Screenshot 2022-11-11 094721.png
 
Postgres Vacuum definitely needs to be looked at and I am working with support to figure out how to have this run automatically (and why it is not). 
 
Hope this helps!!  🍺
 
Doug

victor_menezes
Path Finder

Hey @dbkuhn ,

There is any change of sharing this query? 

Also, how did you cleaned up this indicator you mentioned?


I tried to do the cleanup via delete_containers.py and when I started vacuum process if filled up the disk as it needs kinda of the same amount of data to dispose in disk to process, so it was worse than I thought and now my dev environment is down with 20kb free space in the mount.
I'll have to struggle here to try to expand the mount first in order to be able to re-start the db and start the vacuum again.

0 Karma

dbkuhn
Explorer

Hi Victor,

For the indicators I used the following (was recommended I did it in increments as it can be a lot, as I mentioned)...

For example:

$PHANTOM_HOME/bin/phenv python $PHANTOM_HOME/bin/delete_indicators.pyc --dry-run --before "2022-06-01"
 
(Dry-run function will allow you to see what will be deleted when the command is run, remove it when ready to delete)
 
I recommend having space added to the FS as the delete commands do temporarily use some space, to which I think are temporary DB queries. 
 
My main issue was with the log files in $PHANTOM_HOME/data/db/pg_log/   Once the clean ups completed (I still have a vacuum running on the 'indicator_artifact_record' (a few days now, due to its size) these logs reduced dramatically in size (I'd take a look at those too).  
 
I can not attach a text file to this post. So see below on the query support had me run:

 

WITH constants AS (
    -- define some constants for sizes of things
    -- for reference down the query and easy maintenance
    SELECT current_setting('block_size')::numeric AS bs, 23 AS hdr, 8 AS ma
),
no_stats AS (
    -- screen out table who have attributes
    -- which dont have stats, such as JSON
    SELECT table_schema, table_name, 
        n_live_tup::numeric as est_rows,
        pg_table_size(relid)::numeric as table_size
    FROM information_schema.columns
        JOIN pg_stat_user_tables as psut
           ON table_schema = psut.schemaname
           AND table_name = psut.relname
        LEFT OUTER JOIN pg_stats
        ON table_schema = pg_stats.schemaname
            AND table_name = pg_stats.tablename
            AND column_name = attname 
    WHERE attname IS NULL
        AND table_schema NOT IN ('pg_catalog', 'information_schema')
    GROUP BY table_schema, table_name, relid, n_live_tup
),
null_headers AS (
    -- calculate null header sizes
    -- omitting tables which dont have complete stats
    -- and attributes which aren't visible
    SELECT
        hdr+1+(sum(case when null_frac <> 0 THEN 1 else 0 END)/8) as nullhdr,
        SUM((1-null_frac)*avg_width) as datawidth,
        MAX(null_frac) as maxfracsum,
        schemaname,
        tablename,
        hdr, ma, bs
    FROM pg_stats CROSS JOIN constants
        LEFT OUTER JOIN no_stats
            ON schemaname = no_stats.table_schema
            AND tablename = no_stats.table_name
    WHERE schemaname NOT IN ('pg_catalog', 'information_schema')
        AND no_stats.table_name IS NULL
        AND EXISTS ( SELECT 1
            FROM information_schema.columns
                WHERE schemaname = columns.table_schema
                    AND tablename = columns.table_name )
    GROUP BY schemaname, tablename, hdr, ma, bs
),
data_headers AS (
    -- estimate header and row size
    SELECT
        ma, bs, hdr, schemaname, tablename,
        (datawidth+(hdr+ma-(case when hdr%ma=0 THEN ma ELSE hdr%ma END)))::numeric AS datahdr,
        (maxfracsum*(nullhdr+ma-(case when nullhdr%ma=0 THEN ma ELSE nullhdr%ma END))) AS nullhdr2
    FROM null_headers
),
table_estimates AS (
    -- make estimates of how large the table should be
    -- based on row and page size
    SELECT schemaname, tablename, bs,
        reltuples::numeric as est_rows, relpages * bs as table_bytes,
    CEIL((reltuples*
            (datahdr + nullhdr2 + 4 + ma -
                (CASE WHEN datahdr%ma=0
                    THEN ma ELSE datahdr%ma END)
                )/(bs-20))) * bs AS expected_bytes,
        reltoastrelid
    FROM data_headers
        JOIN pg_class ON tablename = relname
        JOIN pg_namespace ON relnamespace = pg_namespace.oid
            AND schemaname = nspname
    WHERE pg_class.relkind = 'r'
),
estimates_with_toast AS (
    -- add in estimated TOAST table sizes
    -- estimate based on 4 toast tuples per page because we dont have 
    -- anything better.  also append the no_data tables
    SELECT now(), schemaname, tablename, 
        TRUE as can_estimate,
        est_rows,
        table_bytes + ( coalesce(toast.relpages, 0) * bs ) as table_bytes,
        expected_bytes + ( ceil( coalesce(toast.reltuples, 0) / 4 ) * bs ) as expected_bytes
    FROM table_estimates LEFT OUTER JOIN pg_class as toast
        ON table_estimates.reltoastrelid = toast.oid
            AND toast.relkind = 't'
),
table_estimates_plus AS (
-- add some extra metadata to the table data
-- and calculations to be reused
-- including whether we cant estimate it
-- or whether we think it might be compressed
    SELECT current_database() as databasename,
            schemaname, tablename, can_estimate, 
            est_rows,
            CASE WHEN table_bytes > 0
                THEN table_bytes::NUMERIC
                ELSE NULL::NUMERIC END
                AS table_bytes,
            CASE WHEN expected_bytes > 0 
                THEN expected_bytes::NUMERIC
                ELSE NULL::NUMERIC END
                    AS expected_bytes,
            CASE WHEN expected_bytes > 0 AND table_bytes > 0
                AND expected_bytes <= table_bytes
                THEN (table_bytes - expected_bytes)::NUMERIC
                ELSE 0::NUMERIC END AS bloat_bytes
    FROM estimates_with_toast
    UNION ALL
    SELECT current_database() as databasename, 
        table_schema, table_name, FALSE, 
        est_rows, table_size,
        NULL::NUMERIC, NULL::NUMERIC
    FROM no_stats
),
bloat_data AS (
    -- do final math calculations and formatting
    select current_database() as databasename,
        schemaname, tablename, can_estimate, 
        table_bytes, round(table_bytes/(1024^2)::NUMERIC,3) as table_mb,
        expected_bytes, round(expected_bytes/(1024^2)::NUMERIC,3) as expected_mb,
        round(bloat_bytes*100/table_bytes) as pct_bloat,
        round(bloat_bytes/(1024::NUMERIC^2),2) as mb_bloat,
        table_bytes, expected_bytes, est_rows
    FROM table_estimates_plus
)
-- filter output for bloated tables
SELECT now(), databasename, schemaname, tablename,
    can_estimate,
    est_rows,
    pct_bloat, mb_bloat,
    table_mb
FROM bloat_data
-- this where clause defines which tables actually appear
-- in the bloat chart
-- example below filters for tables which are either 50%
-- bloated and more than 20mb in size, or more than 25%
-- bloated and more than 1GB in size
WHERE ( pct_bloat >= 0 AND mb_bloat >= 0 )
    OR ( pct_bloat >= 0 AND mb_bloat >= 0 )
ORDER BY pct_bloat DESC;

 

Tags (1)

victor_menezes
Path Finder

Hey @ouss ,
Sorry to revive this but did you find anything? I'm going thru the same situation here, no luck on finding answers yet.

0 Karma
Get Updates on the Splunk Community!

New Case Study Shows the Value of Partnering with Splunk Academic Alliance

The University of Nevada, Las Vegas (UNLV) is another premier research institution helping to shape the next ...

How to Monitor Google Kubernetes Engine (GKE)

We’ve looked at how to integrate Kubernetes environments with Splunk Observability Cloud, but what about ...

Index This | How can you make 45 using only 4?

October 2024 Edition Hayyy Splunk Education Enthusiasts and the Eternally Curious!  We’re back with this ...