Multiple sourcetypes combine datasets similar to concept Index-to-Match

[TangSauce]

Hello All,

I have been scouring the community and other boards but for the life of me cannot create a SPL query to get the results I need. Below is what I am trying to accomplish, any direction and help would be greatly appreciated.

I have 2 sourcetypes and 1 lookup table.

• SourceA has fields (ID_a, name, date, trimmed_name, env, etc..)
• SourceB has fields (ID, title_id, title, description solution, etc...)
• Lookup has fields (trimmed_name, department, etc...)
• SourceA gets updated weekly, therefore in the existing query I have earliest=-7d to exclude previous data
• Characteristics of SourceA: name is the only unique field, ID_a will be duplicated across name depending on date field, env is duplicated across the dataset, trimmed_name is a field created from trimming name
• SourceA ID_a and SourceB ID are the common field between the two datasets
• Characteristics of SourceB: there are duplicates of ID, title_id, title, description, solution. When deduping title_id, values of title, description, solution, etc... are dedup'd and become unique. This is why I called SourceB a knowledge base. We can dedup SourceB down to the fields above to get a finite list.
• SourceB will have multiple values of ID. But we only need to return fields\values where the dedup'd SourceA ID_a exists. This means from 100 events in SourceA, dedup'd by ID_a, results in 2 ID_a. These 2 ID_a values is what we to find in Source B and return title_id, title, description solution values for.
• At a minimum, final results needed per source are: 
  • SourceA: name, date, env
  • SourceB: title_id, title, description, solution (other fields can be omitted, assuming we can add specific fields back as needed)
  • Lookup: department (other fields can be omitted, assuming we can add specific fields back as needed)

SourceA:

ID_a	name	trimmed_name	date
ABC	ABC_n	AB	01/01/2023
ABC	ABC_n1	AB	01/02/2023
ABC	ABC_n2	AB	01/03/2023
XYZ	XYZ_n	XY	02/01/2023
XYZ	XYZ_n1	XY	02/02/2023

SourceB:

ID	title_id	title	description	solution
ABC	12345	ABC_t	ABC_d	ABC_s
ABC	23456	ABC_t1	ABC_d1	ABC_s1
ABC	93648	ABC_t2	ABC_d2	ABC_s2
XYZ	23456	XYZ_t	XYZ_d	XYZ_s
XYZ	38840	XYZ_t1	XYZ_d1	XYZ_s1
MNO	43245	MNO_t	MNO_d	MNO_s
MNO	36485	MNO_t1	MNO_d1	MNO_s1
RST	84678	RST_t	RST_d	RST_s

Lookup Table:

trimmed_name	department
AB	ABC_dep
XY	XYZ_dep
MN	MNO_dep

Intended Result:

sourceA.name	sourceB.title_id	sourceB.title	sourceB.description	sourceB.solution	sourceA.date	lookup.department
ABC_n	12345	ABC_t	ABC_d	ABC_s	01/01/2023	ABC_dep
ABC_n	23456	ABC_t1	ABC_d1	ABC_s1	01/01/2023	ABC_dep
ABC_n	93648	ABC_t2	ABC_d2	ABC_s2	01/01/2023	ABC_dep
ABC_n1	12345	ABC_t	ABC_d	ABC_s	01/02/2023	ABC_dep
ABC_n1	23456	ABC_t1	ABC_d1	ABC_s1	01/02/2023	ABC_dep
ABC_n1	93648	ABC_t2	ABC_d2	ABC_s2	01/02/2023	ABC_dep
ABC_n2	12345	ABC_t	ABC_d	ABC_s	01/03/2023	ABC_dep
ABC_n2	23456	ABC_t1	ABC_d1	ABC_s1	01/03/2023	ABC_dep
ABC_n2	93648	ABC_t2	ABC_d2	ABC_s2	01/03/2023	ABC_dep
XYZ_n	23456	XYZ_t	XYZ_d	XYZ_s	02/01/2023	XYZ_dep
XYZ_n	38840	XYZ_t1	XYZ_d1	XYZ_s1	02/01/2023	XYZ_dep
XYZ_n	23456	XYZ_t	XYZ_d	XYZ_s	02/02/2023	XYZ_dep
XYZ_n	38840	XYZ_t1	XYZ_d1	XYZ_s1	02/02/2023	XYZ_dep

Query, I have been playing with... From what I gather, I have events for the query but nothing displays when I try to get data from SourceB and SourceA at the same time.

index="myindex" (sourcetype="SourceB" type=INFO) OR (sourcetype="SourceA" date<2023-02-14 env=envA)
| rename ID_a AS ID
| fields ID, name, title, description, solution, date, env
| stats count values(*) AS * values(date) values(title) values(env) BY name
| where count=1

[ITWhisperer]

| eval ID=coalesce(ID, ID_a)
| fields ID, name, title_id, title, description, solution, date, env
| stats list(*) AS * BY ID
| where env="envA"
| eval repA=mvrange(0,mvcount(name))
| mvexpand repA
| foreach name date env
    [| eval <<FIELD>>=mvindex(<<FIELD>>, repA)]
| eval repB=mvrange(0,mvcount(title_id))
| mvexpand repB
| foreach title_id title description solution
    [| eval <<FIELD>>=mvindex(<<FIELD>>, repB)]

[yuanliu]

Problems in your first search:

1. When you rename ID_a as ID, you also erase any value of ID that comes from sourceB.  You need to pull value discretely.
2. Field name only exists in SourceA.  The stats command's groupby will exclude any data from SourceB.
3. The last filter will further exclude any "name" that appear in more than one event in SourceA, which is very unlikely.

Based on your description, you want to

• pull all fields of interest in SourceA;
• add fields of interest in SourceB if ID_a in the former matches ID in the latter; and
• pull additional fields from lookup if trimmed_name in SourceA is found in the lookup.

Combining these, the search should be something like

index="myindex" (sourcetype="SourceB" type=INFO) OR (sourcetype="SourceA" date<2023-02-14 env=envA)
| eval ID = if(sourcetype == "SourceA", ID_a, ID) ``` use ID as universal field name ```
| fields ID, name, title, description, solution, date, env
| stats count values(*) AS * BY ID ``` not by name ```
| lookup lookuptable.csv trimmed_name ``` match trimmed_name ```

Here, I assume that you didn't intend to find singular ID.  But if you need any real restriction on count, you can add it.

[TangSauce]

Based on your description, you want to

• pull all fields of interest in SourceA;
• add fields of interest in SourceB if ID_a in the former matches ID in the latter; and
• pull additional fields from lookup if trimmed_name in SourceA is found in the lookup.

---

The above is accurate. However, the sample query you provide is still returning data from SourceB only.

Apologizes to all, but I may not be describing the problem and expected results inaccurately. Appreciate all the help so far nonetheless.

I updated my original post to provide more clarity.

[yuanliu]

I forgot to filter out data contributed from SourceB that had no match in SourceA.

index="myindex" (sourcetype="SourceB" type=INFO) OR (sourcetype="SourceA" date<2023-02-14 env=envA)
| eval ID = if(sourcetype == "SourceA", ID_a, ID) ``` use ID as universal field name ```
| fields ID, name, title, description, solution, date, env, sourcetype
| stats count values(*) AS * BY ID ``` not by name ```
| where sourcetype == "SourceA" ``` side effect of SPL's multivalue equality ```
| lookup lookuptable.csv trimmed_name ``` match trimmed_name ```

The where filter should be mvcount(sourcetype) > 1 OR sourcetype == "SourceA"; but SPL allows equality operator to match any one of multiple values if one side of the operator is a single value.

[Tom_Lundie]

It looks like you're thinking along the right track.

The common property between sourceA and sourceB is "ID" so that is what we should group by.

Try the following:

index="myindex" (sourcetype="SourceB" type=INFO) OR (sourcetype="SourceA" date<2023-02-14 env=envA)
| eval drop=if(sourcetype="SourceA", 1, null())
| eval ID = coalesce(ID, ID_a)
| table ID, name, title, description, solution, date, env
| eventstats latest(name) as name, latest(trimmed_name) as trimmed_name, latest(date) as date by ID
| where isnull(drop)
| fields - drop
| lookup lookuptable.csv trimmed_name

 
Essentially, we're using sourceB as our main dataset, extracting what we need from sourceA and then dropping the sourceA events from the search results. Then finally it looks like we need to lookup using trimmed_name from your example data.

[TangSauce]

Thanks for the quick reply! Unfortunately, looks like the query is still only returning data from one source.  I played with the query to simplify to just returning 2 data points from SourceA and SourceB, but was unable get it to work.

• The original query shows no results
• Removing the following starts to show results 

  | where isnotnull(drop)
  | fields - drop

Here is an example of what is being returned.

sourceA.name	sourceB.title_id	sourceB.title	sourceB.description	sourceB.solution	sourceA.date	lookup.department
	12345	ABC_t	ABC_d	ABC_s
	23456	ABC_t1	ABC_d1	ABC_s1
	93648	ABC_t2	ABC_d2	ABC_s2
	23456	XYZ_t	XYZ_d	XYZ_s

[Tom_Lundie]

Hi TangSauce,

You're welcome but apologies, I had a typo in my example search, that should have been:

| where isnull(drop)
| fields - drop

I have updated my original post to reflect that.

That should be working now, but your search results are not looking promising...

I've just noticed that you have a date parameter for sourceA are you trying to search different timeranges between the two sources?

index="myindex" (sourcetype="SourceB" type=INFO) OR (sourcetype="SourceA" date<2023-02-14 env=envA)

If date field is being extracted as the _time field for SourceA, Splunk can do the heavy lifting for your filtering:

index="myindex" (sourcetype="SourceB" type=INFO) OR (sourcetype="SourceA" latest="02/14/2023:00:00:00" env=envA)

Otherwise you'll need to manually filter the dates out yourself:

index="myindex" (sourcetype="SourceB" type=INFO) OR (sourcetype="SourceA" env=envA)
| where sourcetype="SourceB" OR (strptime(date, "%Y-%m-%d") > strptime("2023-02-14", "%Y-%m-%d"))

Either way, focus on writing a search that gets the correct data that you require from SourceA and SourceB together. Make sure that data from both SourceA and SourceB are present with the first section of the search.

Then you can use the rest of the search from my original answer to achieve the overall goal.

| eval drop=if(sourcetype="SourceA", 1, null())
| eval ID = coalesce(ID, ID_a)
| table ID, name, title, description, solution, date, env
| eventstats latest(name) as name, latest(trimmed_name) as trimmed_name, latest(date) as date by ID
| where isnull(drop)
| fields - drop
| lookup lookuptable.csv trimmed_name

[TangSauce]

Thanks again for the help. Still struggling and trying different ways to get the search to return or at least include the dataset and fields I need. Again, running into only being able to show values from SourceB but not SourceA. The date field in SourceA does not equal _time, but is specific to when the name field was changed (1-1 relationship).

I assume I would need to confirm the returned results are from both SourceA and SourceB via Events (~7.9k events). I have been able to play with python to achieve the goal but, not everyone in my org will be able access this data via a spreadsheet.