[BRC-2905] Feed back PS-detected data corruption signals to SK and PG… (#12748)

… walproposer (#895)

Data corruptions are typically detected on the pageserver side when it
replays WAL records. However, since PS doesn't synchronously replay WAL
records as they are being ingested through safekeepers, we need some
extra plumbing to feed information about pageserver-detected corruptions
during compaction (and/or WAL redo in general) back to SK and PG for
proper action.

We don't yet know what actions PG/SK should take upon receiving the
signal, but we should have the detection and feedback in place.

Add an extra `corruption_detected` field to the `PageserverFeedback`
message that is sent from PS -> SK -> PG. It's a boolean value that is
set to true when PS detects a "critical error" that signals data
corruption, and it's sent in all `PageserverFeedback` messages. Upon
receiving this signal, the safekeeper raises a
`safekeeper_ps_corruption_detected` gauge metric (value set to 1). The
safekeeper then forwards this signal to PG where a
`ps_corruption_detected` gauge metric (value also set to 1) is raised in
the `neon_perf_counters` view.

Added an integration test in
`test_compaction.py::test_ps_corruption_detection_feedback` that
confirms that the safekeeper and PG can receive the data corruption
signal in the `PageserverFeedback` message in a simulated data
corruption.

## Problem

## Summary of changes

---------

Co-authored-by: William Huang <william.huang@databricks.com>
This commit is contained in:
Suhas Thalanki
2025-07-29 16:40:07 -04:00
committed by GitHub
parent 7cd0066212
commit 07c3cfd2a0
18 changed files with 208 additions and 1 deletions

View File

@@ -45,6 +45,7 @@ DatabricksMetricsShmemInit(void)
pg_atomic_init_u32(&databricks_metrics_shared->index_corruption_count, 0);
pg_atomic_init_u32(&databricks_metrics_shared->data_corruption_count, 0);
pg_atomic_init_u32(&databricks_metrics_shared->internal_error_count, 0);
pg_atomic_init_u32(&databricks_metrics_shared->ps_corruption_detected, 0);
}
}
/* END_HADRON */
@@ -440,6 +441,7 @@ neon_get_perf_counters(PG_FUNCTION_ARGS)
{"sql_index_corruption_count", false, 0, (double) pg_atomic_read_u32(&databricks_metrics_shared->index_corruption_count)},
{"sql_data_corruption_count", false, 0, (double) pg_atomic_read_u32(&databricks_metrics_shared->data_corruption_count)},
{"sql_internal_error_count", false, 0, (double) pg_atomic_read_u32(&databricks_metrics_shared->internal_error_count)},
{"ps_corruption_detected", false, 0, (double) pg_atomic_read_u32(&databricks_metrics_shared->ps_corruption_detected)},
{NULL, false, 0, 0},
};
for (int i = 0; databricks_metrics[i].name != NULL; i++)

View File

@@ -183,6 +183,7 @@ typedef struct
pg_atomic_uint32 index_corruption_count;
pg_atomic_uint32 data_corruption_count;
pg_atomic_uint32 internal_error_count;
pg_atomic_uint32 ps_corruption_detected;
} databricks_metrics;
extern databricks_metrics *databricks_metrics_shared;

View File

@@ -1887,6 +1887,12 @@ ParsePageserverFeedbackMessage(WalProposer *wp, StringInfo reply_message, Pagese
ps_feedback->shard_number = pq_getmsgint(reply_message, sizeof(uint32));
psfeedback_log("%u", key, ps_feedback->shard_number);
}
else if (strcmp(key, "corruption_detected") == 0)
{
Assert(value_len == 1);
ps_feedback->corruption_detected = pq_getmsgbyte(reply_message) != 0;
psfeedback_log("%s", key, ps_feedback->corruption_detected ? "true" : "false");
}
else
{
/*

View File

@@ -374,6 +374,8 @@ typedef struct PageserverFeedback
XLogRecPtr remote_consistent_lsn;
TimestampTz replytime;
uint32 shard_number;
/* true if the pageserver has detected data corruption in the timeline */
bool corruption_detected;
} PageserverFeedback;
/* BEGIN_HADRON */

View File

@@ -49,6 +49,7 @@
#include "libpqwalproposer.h"
#include "neon.h"
#include "neon_perf_counters.h"
#include "neon_walreader.h"
#include "walproposer.h"
@@ -741,6 +742,11 @@ record_pageserver_feedback(PageserverFeedback *ps_feedback, shardno_t num_shards
Assert(ps_feedback->shard_number < MAX_SHARDS);
Assert(ps_feedback->shard_number < num_shards);
// Begin Hadron: Record any corruption signal from the pageserver first.
if (ps_feedback->corruption_detected) {
pg_atomic_write_u32(&databricks_metrics_shared->ps_corruption_detected, 1);
}
SpinLockAcquire(&walprop_shared->mutex);
// Hadron: Update the num_shards from the source-of-truth (shard map) lazily when we receive