Client approves a post in the Sheet. Then someone fixes a typo or swaps the image, and n8n publishes something nobody approved.
Fix: at approval, hash caption + media URL. Recheck before publishing. Mismatch = back to pending.
Free 4-page sample: https://t.co/kj1gHAjehQ
@syifa_anns Nice one. A guard worth adding: skip comments where https://t.co/usuJ9pAv38 is your own Page, or each auto-reply becomes a new comment event and the flow answers itself. Dedupe on comment_id too, since Meta can resend the webhook.
@quietpatterns88 Empty comment list is a classic. One common cause: while your Meta app is in Development mode, the API only returns comments from accounts with a role on the app. Real users' comments show up once it's Live with instagram_manage_comments approved.
Scheduled Instagram posts going out an hour off after a clock change?
Your Sheet says "10:00" but never says which 10:00.
Fix:
- Store publish_at in UTC, plus the zone it was meant for
- Set the workflow timezone in n8n, don't trust the server default
- Compare in UTC only
@AdepetuTayo What helped us with the screencasts: one short video per permission, on a test account, showing the exact screen where the user grants it and the feature that uses it. And request only the permissions you actually call, fewer things to prove.
@MartinSzreter Yes, and give 'publishing' a timeout. If a row sits there over ~15 min, the run probably died mid-call: don't auto-retry. Check the container's status_code first. PUBLISHED means mark it done, anything else goes to a human.
n8n flow posted to Instagram fine for 2 months, then every run fails with error 190?
That's the long-lived token hitting its ~60-day expiry.
Fix:
- Refresh it on a schedule (every ~30 days), not when it breaks
- Store the expiry date next to the token
- Alert on 190, no retries
@pendpost Useful numbers. For long reels, try the video_url path instead: host the file on a public URL (R2/S3), let Meta fetch it, then poll status_code until FINISHED. The slow transfer stops racing your request timeout.
@syifa_anns Nice setup. One thing worth adding: write the approval state back to the queue row (pending > approved > published). Then a double tap in Telegram or a late approval can't publish twice or post something stale. An expiry on pending approvals helps too.
Before your n8n flow publishes to Instagram, check 4 things:
1. Media URL signed just now, not hours ago
2. Image is JPEG, not PNG/WebP
3. Row flipped to "publishing" so a rerun skips it
4. Token not close to expiry
Free 4-page checklist: https://t.co/kj1gHAjehQ
@CraigV7912 Two layers for us: an Error Trigger workflow for crashes, plus a daily check that each workflow actually produced its expected output (e.g. one post per scheduled row). The second one catches the runs that stay green but did nothing.
Real bug: the same Instagram post went out twice.
media_publish timed out, Retry On Fail ran it again, but the first call had worked.
Fix: save the container id in your queue first. On retry, check its status_code. PUBLISHED means done, skip publish.
@PlugiversePower Also worth answering Meta first: put a Respond to Webhook node right after the trigger so it gets its 200 in milliseconds, then run the AI step. Fewer retries to dedupe in the first place.
@LeadPulseHQ Good catch. Same thing happens with manual reruns from the execution log: they reuse the old URL. Signing it inside the publish branch, right before the Instagram call, fixes both.
Instagram says "media could not be fetched" in your n8n flow? Check before touching the workflow:
1. URL is public, no login or redirect
2. Signed R2/S3 link hasn't expired (give it hours, not minutes)
3. Image is JPEG, not WebP or PNG
4. Aspect ratio is between 4:5 and 1.91:1
Instagram Reel stuck in IN_PROGRESS in your n8n flow?
Poll the container's status_code every ~30s, with a cap (say 10 tries).
FINISHED: publish.
ERROR or still stuck after the cap: mark the row failed, alert a human.
One stuck container shouldn't block the rest of the queue.
@techie_piyush@Meta@instagram Worth checking the X-App-Usage and X-Business-Use-Case-Usage response headers first, they show which limit you're actually hitting. Content publishing also has its own 24h cap per account, separate from the hourly call limit.
@AlekVectis@NathanHChan Been there. What stopped it for us: swap to a long-lived token (~60 days), refresh it on a schedule well before expiry, and alert on error 190. Or use a system user token if it's a business setup.
@k4k0ch What works for us: a dry_run flag that runs the full flow (read queue, pick due row, build caption) and returns the plan without calling Meta. Then one live run on a private test account. Creating an IG container without media_publish is also a safe rehearsal, it never goes live.
@elyasadev Good call. I'd also log the original format when converting, so if a post looks soft later you know it was the re-encode and not Instagram's compression.