What causes "Cannot fetch from cache %s" in prestodb/presto?

Look at the cause (e.getCause()) logged with this exception to find the real failure

prestodb/presto · error · PinotException

PINOT_UNCLASSIFIED_ERROR

PINOT_UNCLASSIFIED_ERROR

Error message

Cannot fetch from cache %s

What it means

PinotConnection.getFromCache wraps any ExecutionException from a Guava loading cache (all-tables or per-table columns) into PINOT_UNCLASSIFIED_ERROR. The real failure is whatever caused the underlying cache load to fail, attached as the cause.

Source

Thrown at presto-pinot-toolkit/src/main/java/com/facebook/presto/pinot/PinotConnection.java:87

                                        nullHandlingEnabled,
                                        pinotConfig.isCaseSensitiveNameMatchingEnabled());
                            }
                        }, executor));

        executor.execute(() -> this.allTablesCache.refresh(ALL_TABLES_CACHE_KEY));
    }

    private static <K, V> V getFromCache(LoadingCache<K, V> cache, K key)
    {
        V value = cache.getIfPresent(key);
        if (value != null) {
            return value;
        }
        try {
            return cache.get(key);
        }
        catch (ExecutionException e) {
            throw new PinotException(PINOT_UNCLASSIFIED_ERROR, Optional.empty(), "Cannot fetch from cache " + key, e.getCause());
        }
    }

    public List<String> getTableNames()
    {
        return getFromCache(allTablesCache, ALL_TABLES_CACHE_KEY);
    }

    public PinotTable getTable(String tableName)
    {
        List<PinotColumn> columns = getPinotColumnsForTable(tableName);
        return new PinotTable(tableName, columns);
    }

    private List<PinotColumn> getPinotColumnsForTable(String tableName)
    {
        return getFromCache(pinotTableColumnCache, tableName);
    }

View on GitHub (pinned to 55bb57d202)

Solutions

Look at the cause (e.getCause()) logged with this exception to find the real failure
Verify Pinot controller connectivity with curl and restore connectivity if down
Retry after the controller recovers; the cache load is not permanently poisoned
Pin/validate controller and connector versions if the cause is a JSON parse error

Defensive patterns

Strategy: retry

Validate before calling

// Preflight controller check before metadata reads
curl -sf http://<controller>:9000/tables || echo 'controller failing'

Try / catch

try {
    List<String> tables = connection.getTableNames();
} catch (PinotException e) {
    if (e.getMessage().startsWith("Cannot fetch from cache")) {
        // inspect e.getCause(); usually controller/network - retry after recovery
    }
    throw e;
}

Prevention

Always log e.getCause() for PINOT_UNCLASSIFIED_ERROR — it carries the real failure
Monitor controller availability from Presto nodes
Align controller/connector versions to avoid schema JSON parse breaks
Add modest cache refresh throttling to avoid controller hammering

When it happens

Trigger: getTableNames() or getPinotColumnsForTable() triggers a cache load that throws a non-PinotException — usually a controller HTTP error or JSON parse failure while listing tables or fetching schema.

Common situations: Controller temporarily unreachable while metadata is being read; controller returns malformed schema JSON after an upgrade; concurrent cache loads hitting controller rate limits.

Related errors

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/8951870583d9ffbd. Report an issue: GitHub.