Skip to main content

Activity log to S3

Delegaris can copy everything in Activity to an S3 bucket every 15 minutes, as JSON Lines. Splunk, Microsoft Sentinel, Elastic and most SIEMs can read a bucket like this directly. It also keeps a copy of the log outside Delegaris.

Administrators only

The export is set up under Settings > Health and backups, which only administrators can open.

Before you start​

The export uses the service's own AWS identity (see Service identity), not a role in a member account. That identity needs s3:PutObject on the bucket.

If the service runs on EC2 with the instance role from iam/hub-role.yaml, redeploy that stack with the bucket name in its AuditBucketName parameter; it adds the permission. Otherwise, add a statement like this to whatever identity the service uses:

{
"Effect": "Allow",
"Action": "s3:PutObject",
"Resource": "arn:aws:s3:::corp-security-logs/*"
}

Files are written with S3-managed encryption (SSE-S3).

Turning it on​

  1. Open Settings > Health and backups and find Activity log to S3.
  2. Tick Copy the activity log to S3.
  3. Enter the Bucket, for example corp-security-logs.
  4. Optionally change the Folder (prefix). The default is delegaris/audit/; files go under it by date.
  5. Press Save.

To check it straight away, press Export now. A message says what was written, or why it failed. Under the form, a line shows the last export: when it ran, how many entries have been exported so far and the latest file, or the error.

The files​

Each export writes one file (or more, for a large backlog) named:

<prefix>YYYY/MM/DD/<time>-<first id>-<last id>.jsonl

For example delegaris/audit/2026/11/13/20261113T091500Z-4211-4268.jsonl. The date and time are when the export ran, in UTC; the two numbers are the first and last entry in the file.

Each line is one entry, as JSON:

{"id":4212,"time":"2026-11-13T09:02:41Z","user":"jsmith","action":"keep_on","ok":true,"account_id":"111111111111","region":"eu-west-2","resource_type":"ec2","resource_id":"i-0abc123def4567890","detail":"until 2026-11-13T22:00:00Z; reason: release testing","error":null,"ip":"10.1.2.3","source":"delegaris"}
FieldMeaning
idThe entry's number, increasing
timeWhen it happened, in UTC
userWho did it
actionWhat was done, by its internal name, such as start, keep_on, drop_tag or login
oktrue if it worked, false if it was refused or failed
account_id, region, resource_type, resource_idThe resource, where there is one
detailAnything else recorded, as text
errorWhy it was refused or failed
ipThe address it came from
sourceAlways delegaris

How it behaves​

  • The first export includes everything already recorded. After that, only what is new.
  • Nothing is skipped or sent twice. A failed upload is retried from the same place on the next pass.
  • A large backlog is sent in files of up to 5,000 entries, spread over several passes.
  • With more than one copy of the service, only the copy doing the timed work runs the 15-minute export. See High availability.