Email puller says “every minute”, but emails are processed in ~30‑minute batches. Is this expected?

[ttang]

Hi Atlassian team,

We are using Jira Service Management Data Center (Jira 9.16.x, JSM 5.x) with an Email Channel connected to Microsoft 365 via Microsoft Graph (OAuth 2.0).

According to the official documentation:

“Email puller: connects to your mail servers every minute and pulls the email data into the database.”

However, in our production environment we consistently observe the following behavior:

• Emails are successfully delivered to the Microsoft 365 mailbox immediately 

• But JSM processes multiple new emails together in the same timestamp 

• The processing time is often ~20–30 minutes after the email was sent 

• This happens for different users in different regions (Europe, Brazil), and the emails are always processed in the same batch

Example:

– Email sent at 09:21 and 09:31 (local sender times)

– Both requests are created at exactly 15:56 (JSM email audit log time)

This behavior is very consistent and reproducible.

We understand that JSM uses a polling (pull) mechanism. Our question is not about troubleshooting, but about clarification:

1) Does “every minute” in the documentation refer only to the scheduler job trigger, rather than a real-time guarantee?

2) Is it expected in JSM Data Center that actual successful email fetching can happen in larger effective intervals (for example due to OAuth throttling, mail provider limits, or internal scheduling)?

3) Is there any official SLA or expectation documented by Atlassian regarding email-to-request creation latency?

We want to confirm whether this behavior is considered expected by design.

Thanks in advance.

[Arkadiusz Wroblewski]

Hello and welcome to the Community @ttang 

it’s better to read that "every minute" in the docs as the scheduler’s heartbeat rather than a real-time guarantee. In JSM Data Center, processing is a two-step dance where the Puller fetches the mail and the Processor handles the ticket creation. Even in a perfect setup, Atlassian suggests this takes at least two minutes end-to-end. While there isn't an official SLA, a 20-30 minute delay definitely isn't "normal" or documented as intended behavior.

Keep in mind that your logs show when Jira finally processed the email, not when it actually arrived in the mailbox, so you are looking at internal Jira latency. I’d take a look at your Scheduler Details for the Global Puller and Processor to see if they are skipping cycles or stalling. If you need a formal statement on whether that kind of lag is officially "acceptable," you'll likely have to reach out to Atlassian Support, as the community can only help interpret how the engine is supposed to run.

.