Use this page when a Lake table is deployed and you need to load it now, stop a running load, empty or remove a table, or find out why a load failed. Table actions are on the Settings tab of the table popup (see Configure a table). Datasource actions are in the Lake tree.
The date you pick is the date the load runs for, and the Sync Log records it as Current Date. One load runs at a time for a table, so its actions stay disabled until the run finishes, and a run appears in the log as soon as it starts. While a job runs, the Storage card shows a Syncing badge and refreshes about every 5 seconds; a manual Refresh confirms with Table statistics refreshed. If the sync does not start, the dialog shows Failed to synchronize. Please try again.
The Sync Log lists every load of the table, newest first, below the Storage card. A finished run keeps its final state, so you can open it later and still see the result.
| Column | What it shows |
|---|---|
| Log ID | The number of the run. |
| Status | Where the run is. See the statuses below. |
| Note | Extra detail, such as the reason a run was skipped. |
| Start Date and End Date | When the run started and finished. End Date is hidden by default. |
| Duration | How long the run took, or how long it has run so far. |
| Current Date | The date the run loaded for. |
| Processed Rows | The number of rows the run processed. |
| Error Stack | The first line of the error. Click the cell to read the full message. |
| Status | Meaning |
|---|---|
| Pending | The run is queued and has not started. |
| Running | The run is in progress. |
| Finished | The run completed. |
| Error | The run failed. Open Error Stack. |
| Aborted | You stopped the run. |
| Skipped | The run did not load. The Note gives the reason. |
While a run is live, the Sync Log refreshes every few seconds. To refresh it by hand, click the icon with the tooltip Refresh data. An empty log reads No synchronization logs found.
For a file data source, the run shows what happened to the files as extra steps: Files Loaded, Rows Rescued and Files Skipped with the reason, such as no date.
For a Connect Any API load, a run with no error stack opens Runner Diagnostics instead (see Connect any API). If the diagnostics have not been collected yet, the window says so and asks you to try again. When they are collected, the steps show what the load did; parts you cannot access are hidden, and a very long message is cut short with a note.
To remove an entry from the log, click the trash icon on a row that is not running (screen readers name it Delete log entry), then click Confirm in Delete Log Entry.
To stop one load:
The cross appears only on rows that are Pending or Running. Stopping one run ends its query and marks the run Aborted with the note Aborted by user.
To stop the sync of a whole datasource:
Stop Sync restores the data to the state it had before the sync started, so nothing partial is kept. Re-syncing afterwards works normally. The message Sync stopped confirms it, and the datasource then shows a grey Stopped badge; see The Lake screen for what each badge means.
Stop Sync appears only while the datasource shows Processing, and the row keeps that state until the next refresh. If the sync already ended, the request is refused with Data source "…" is not currently syncing; refresh the tree to see the current state.
Sync Schema re-reads the tables and columns from the source and detects newly added tables and columns.
To reload only the schema that Lake shows, without contacting the source, use Refresh in the same menu.
Sync Schema adds only what is missing, so it is safe to run again: once the Lake mirrors the source it changes nothing. When it finishes, the message reads Schema synchronized; if there was nothing to add, it reads Nothing to synchronize. A column whose source type changed is recorded as a suggestion while the Lake keeps its type, and a column that no longer exists in the source is reported and kept.
New columns are added with SY and MI switched off, after any locked columns. A column is loaded with each run only when SY is on. MI controls whether the column is created in the table. Switch on the columns you need and click Apply; the details are on Set up a table's columns. If a later deploy fails, the banner offers Redeploy.
The Storage card on the Settings tab holds the actions for the table's data.
| Button | What it does |
|---|---|
| Synchronize | Starts one load now, for the date you pick. |
| Truncate | Removes the table's data, or only the partitions you choose. |
| Drop | Removes the data from ClickHouse and keeps the table definition, so you can deploy again. |
| Delete | Removes the table from the Lake. |
| Refresh | Reloads the Size, Rows and Partitions figures. |
Truncate, Drop and Delete remove data permanently. Delete also removes the table from the Lake.
To truncate, click Truncate. On a partitioned table, Truncate Partitions opens and loads the partition list (Loading partitions...). Choose the partitions and click Truncate: the dialog shows what you are about to remove (You are about to truncate {N} partitions ({rows} rows) from this table.) and warns This action cannot be undone. The selection resets every time you open the dialog, and if the table has no partitions yet it says This table has no partitions yet - there is nothing to truncate. On any other table, Truncate Table asks you to confirm: This table is not partitioned, so truncating clears all of its data. This action cannot be undone. Unsaved changes on the General step are saved before the truncate runs.
To drop, click Drop, then click Drop table in Drop this table from ClickHouse?. All data is permanently deleted, and the definition stays, so you can deploy again. Anything that depends on the table is not dropped and will break until you fix it or redeploy the table.
To delete, click Delete, then click Delete in Delete Table. The dialog first checks what depends on the table; if something does, the button reads Delete anyway, and What uses a table shows what the check covers. If the delete is refused, the dialog stays open with the reason.
If the table's workflow is scheduled to run within the next hour, a manual sync is refused and the dialog Synchronization blocked by a scheduled run opens. It offers two choices:
To keep the schedule, close the dialog (or click outside it). The sync does not start, and the scheduled run goes ahead. See Schedules and when a workflow starts.
For a Connect Any API table, click the Error Stack cell of the run. A panel above the raw error shows Load failed or Load finished with warnings, with Stream, Mode, Pages, Rows and Elapsed. A run that finished with warnings but no error shows a warning icon in that cell. For every message, see Lake troubleshooting.
If your role lacks the permission, the action is refused. See Who can do what in the Lake.
Pick the date the load should run for. The Sync Log records it as Current Date. The picker opens with the current date and time already set, so change it only when the load needs a different one.
The table's workflow is due within the next hour and the table has Synchronize on, so a manual sync could clash with that run. Choose one option in the dialog, or close it and let the scheduled run go ahead.
One of these is in the way: a run is still going, or the table sits in a workflow whose schedule is about to fire or is running. Hover over the button to read the reason, and follow it: wait for the run to finish, turn off Synchronize on the Table > General tab, or adjust the workflow schedule.
No. Stop Sync restores the data to the state it had before the sync started, and you can sync again afterwards. Stopping one run from the Sync Log ends that run; it is marked Aborted with the note Aborted by user.
On a partitioned table, click Truncate, choose that day's partition dates in Truncate Partitions, and click Truncate. Then click Synchronize with that day's date in Synchronize Table.
Partitions follow the table's sync type (see Choose a sync type). Truncate Partitions is offered only for tables whose sync type produces partitions; for other tables the whole table is truncated instead, and a table with no partition columns has no partitions to target. The dialog says which case you are in.
The button appears only when the table holds data. If its Size, Rows and Partitions figures are all zero, there is nothing to truncate. Drop, in turn, appears only on a deployed table.
Truncate removes the data and keeps the table. Drop removes the data from ClickHouse but keeps the table definition, so you can deploy it again. Delete removes the table from the Lake.
Not yet. New columns start with SY and MI switched off. Switch on the columns you need and click Apply; a column is loaded with each run only when SY is on. See Set up a table's columns.
For a file data source, the run shows what happened to the files: Files Loaded counts the files read, Rows Rescued counts the rows rescued when a value could not be read as its column type, and Files Skipped lists the files left out, each with its reason, such as no date.
Click the red Failed badge to open Error Details, and copy the message if you need help. Fix the cause, then run Sync Schema on the datasource or Synchronize the table again.
Open Job runs. In the Lake, you can also open Workflows and choose Runs. See Workflows in the Lake.