Running into issues with visualizations erroring out and/or not loading relationships defined in the model, so they can't be filtered properly. I've also noticed our model page takes a long time to load.. Could this be related? I opened a ticket about it here - https://support.gooddata.com/hc/en-us/requests/131655

This kind of combination—visualizations intermittently erroring or not loading filter values, plus the LDM/model page being slow to open—does show up in several past cases and can be related, but the exact root cause differs by environment. From the knowledge we have, the main recurring patterns are: 1. Backend / DB performance or connection issues - Slow or failing filter loads and 500 errors often trace back to: - Slow or timing‑out SQL on the data source (GoodData Cloud has a hard 30s SQL timeout). [[Explore timeout](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1721144102677359?thread_ts=1721144102.677359&cid=C04S1MSLEAW)] - DB connection problems such as

Connection is closed

after ~30s, which then cause filters to fail to load. [[Filter connection closed](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1744607227381359?thread_ts=1744607227.381359&cid=C04S1MSLEAW)] - In another case, intermittent filter/visualization failures over ~20 days were ultimately linked to performance issues on the customer’s Postgres side, even though other tools looked fine. [[Intermittent filters](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1747128915431269?thread_ts=1747128915.431269&cid=C04S1MSLEAW)] This can also manifest as the LDM editor “spinning” for a long time or even showing “Connection failed” while loading datasets, especially on large or slower data sources. [[Slow LDM editor](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1741791483762689?thread_ts=1741791483.762689&cid=C04S1MSLEAW)] 2. Query / model complexity and limits - Some visualizations hit platform limits (e.g., too many datapoints, or cross‑tab row limits), which can cause errors or timeouts. [[Limits hit](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1744607227381359?thread_ts=1744607227.381359&cid=C04S1MSLEAW)] - The order and structure of metrics and joins can make a query much heavier; the first metric becomes the “driver” and can lead to timeouts if it forces complex joins. [[Metric order impact](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1761532143785779?thread_ts=1761532143.785779&cid=C04S1MSLEAW)] - Many‑to‑many relationships and inner joins in the LDM can cause filters to return no values if intermediate tables are empty, even though the base data exists. [[M2M filters empty](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1740741477107939?thread_ts=1740741477.107939&cid=C04S1MSLEAW)] 3. Intermittent / “fixes after reload” behavior - Several customers reported filters or visualizations failing until they: - Reloaded the page multiple times, or - Opened the same dashboard/insight directly in GoodData UI (not embedded), after which embedded views started working again. [[Embedded filters intermittent](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1752594898882539?thread_ts=1752594898.882539&cid=C04S1MSLEAW)] - Others saw filters fail right after cache reloads or after source changes, then work intermittently. [[Cache / source change](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1701443681372239?thread_ts=1701443681.372239&cid=C04S1MSLEAW)] 4. Known bugs / driver issues - There have been specific bugs (e.g., Redshift JDBC driver, invalid filter values causing 500s) that produced filter‑load errors until fixed by engineering. [[Redshift filter bug](https://gooddataconnect.slack.com/archives/C04S1MSLEAW/p1722015095455419?thread_ts=1722015095.455419&cid=C04S1MSLEAW)] Because you already opened a support ticket, that’s the right path: support can pull your exact trace IDs and SQL, which I cannot see from here. The Slack‑based cases above show that the real cause is usually only clear once engineers inspect: - Trace IDs from failing visualizations/filters (Network or Console tab). - Whether the issue is workspace‑wide or limited to specific dashboards/insights. - Your LDM structure (joins, many‑to‑many, very large fact tables, etc.). - Data source performance (query times, timeouts, connection settings). [[Performance guidance](https://www.gooddata.com/docs/cloud/connect-data/performance/)] Given the symptoms you describe (visualizations erroring, filters not seeing relationships, slow model page), it is plausible they are related to either: - Data‑source performance / connection behavior, or - LDM complexity / query limits, or a combination of both, but the knowledge sources do not contain enough information to say which one applies to your specific workspace. While support works your ticket, the most useful things you can add there (if you haven’t already) are: 1. A few trace IDs from: - A failing visualization, - A failing filter load, - A slow LDM‑editor load (if any error appears in DevTools). 2. Examples: - One dashboard where relationships/filters don’t behave as expected. - Whether the same issue appears in both embedded and native GoodData UI. 3. Any recent changes: - To your data source (schema, views, indexes, connection timeouts), - To the LDM (new joins, many‑to‑many, large tables added). Support can then correlate your case with the patterns above and confirm whether this is a performance/config issue, a modeling issue, or a product bug.

related - i turned off flex cache recently because it kept delivering stale data to customers.

Hi Matt, lets continue in the ticket

@Julius Kos i updated the ticket to be more clear, and provide a concrete example of the problem.