nano SIEM
Integrations

CrowdStrike FDR

Supported in:Self-hosted

Collect CrowdStrike Falcon Data Replicator telemetry from its SQS queue and S3 objects

Collect Falcon endpoint telemetry (ProcessRollup2, DnsRequest, NetworkConnectIP4 and the rest) from CrowdStrike Falcon Data Replicator (FDR). FDR announces each batch with one SQS message, a manifest that lists gzipped event files in a CrowdStrike-owned S3 bucket. nano reads the manifest, downloads every listed file, and sends each event line to the CrowdStrike Falcon parser.

Setup

You need source_configs:create, and credentials:use to attach a credential. Binding a credential to a new source also needs credentials:edit, as does pointing a source with a credential at a different queue, Region or endpoint. Review the pull permissions matrix before granting access.

1. Get the FDR feed from CrowdStrike

In the Falcon console, open Falcon Data Replicator (under Support and resources) and create a feed. Falcon shows a queue URL, an access key ID and a secret access key; copy all three. CrowdStrike owns the queue and the bucket and grants the key exactly what nano needs: receive, delete and change visibility on the queue, and read on the bucket.

2. Add the credential and the pull source

Open Ingestion → Log Sources → Cloud pull. Under Credentials, add an AWS credential with the FDR access key ID and secret. To read through a role in your own account instead, give the credential an assume-role ARN (and external ID). With no keys, the deployment's own identity assumes that role. Then choose Add cloud pull, pick CrowdStrike FDR, and enter:

FieldValue
FDR SQS Queue URLThe queue URL from Falcon, e.g. https://sqs.us-west-2.amazonaws.com/123456789012/cs-mav-cannon-…-queue
AWS RegionFrom queue URL (default), or the queue's Region
Start fromEverything on the queue (default), Only batches after the first start, or Resume after the last delivered batch
CredentialThe FDR AWS credential
Checksum, File concurrency, Max in-flight batchesLeave the defaults (verify, 4, 64) unless you need to tune memory or throughput

Install the CrowdStrike Falcon (OCSF) parser from the parser repository and bind it to this source. With no routing rules, every FDR event is routed to crowdstrike_falcon, the Falcon parser's source type. Then Deploy the source.

Verify

Batches arrive every few minutes. Search for Falcon process launches:

metadata.log_source=crowdstrike_falcon class_uid=1007 | head 20

stats count by class_name over the last hour shows which event types arrived. Each FDR event type maps to its OCSF class: process activity, network activity, DNS activity, file system activity, authentication and detection findings.

How collection works

  • Held until stored. nano deletes an FDR message only after every event of every file in its batch is accepted by the ClickHouse write buffer. Until then it extends the message's visibility every half visibility timeout, for up to 10 hours.
  • Failures retry the whole batch. A failed download, a checksum mismatch, a corrupt file or a write failure releases the message, and FDR delivers it again about 30 seconds later. After three deliveries nano leaves the message alone for the queue's own retention.
  • Duplicates are possible, loss is not. A retried or restarted batch is read again from the start, so events that were already stored arrive a second time.
  • Restarts never skip unfinished batches. Start from only skips a batch the first time FDR delivers it. A batch nano received earlier and has not finished is always read again.
  • Lookup snapshots are dropped. FDR also delivers host inventory snapshots (fdrv2-*: aidmaster, managedassets) on the same queue. They are not events, so nano acknowledges and discards them rather than sending them to the parser.
  • Memory. Each file (about 30-35 MB compressed) is held in memory while it is verified, so memory grows with File concurrency.

Troubleshoot

Need help with these checks? Contact support with the deployment, request ID and time with timezone.

Access denied in Vector logs

Check that the credential holds the FDR key from the Falcon console, that the queue URL is the feed's, and that the Region matches the queue. Keys expire when the feed is regenerated in Falcon; update the credential in place.

Nothing arrives

FDR delivers a batch every few minutes, not continuously. Check that the source is deployed and that the Falcon parser is bound to it. Check whether another tool also reads the feed: a second consumer takes batches nano never sees.

Events arrive twice

A batch that failed part-way, or was in flight during a restart, is read again in full, so some of its events are stored twice with identical content. This is how FDR guarantees nothing is lost. If duplicates keep appearing, look in Vector logs for the failure that keeps releasing the batch.

What's next

On this page