SFTP: replace a dropped connection; write the index cache in one transaction

A Hetzner Storage Box drops an SFTP connection that carried a lot of data
and then stayed quiet for a few minutes; files_external keeps using the dead
client, on which every file looks missing. StorageBackend now checks the
connection after a minute without requests, and replaces a connection that
is dead after a failed request (once per request; put, move and delete are
not repeated). SftpReadStream closes a connection it left in an unknown state.

The verify cache refresh took more than 6 minutes on PostgreSQL for 27,547
packs, one autocommitted insert per pack: putMany()/replaceAll() write in one
transaction. After a verify the cache holds the index files as they are, so a
blob in two packs stays cached for both.
This commit is contained in:
2026-10-12 01:58:54 +02:00
parent d659322e60
commit 9c06ab2841
9 changed files with 442 additions and 33 deletions
+72 -7
View File
@@ -30,19 +30,84 @@ final class DbIndexCache implements IIndexCache {
}
public function put(string $repositoryId, string $packId, array $entries): void {
try {
$this->insert($repositoryId, $packId, $entries);
} catch (Exception $e) {
if ($e->getReason() !== Exception::REASON_UNIQUE_CONSTRAINT_VIOLATION) {
throw $e;
}
}
}
/**
* In one transaction: on PostgreSQL with its default settings every statement of its own
* waits for a disk sync, which made 27,547 rows take more than 6 minutes. Rows already there
* are replaced. Should the transaction fail (another process writing the same rows), the
* rows are written one by one, keeping rows that are already there.
*/
public function putMany(string $repositoryId, array $packs): void {
if ($packs === []) {
return;
}
try {
$this->transaction(function () use ($repositoryId, $packs): void {
$this->remove($repositoryId, array_keys($packs));
$this->insertAll($repositoryId, $packs);
});
} catch (\Throwable $e) {
if ($this->db->inTransaction()) {
throw $e; // a transaction of the caller's, which may be spoilt now
}
foreach ($packs as $packId => $entries) {
$this->put($repositoryId, $packId, $entries);
}
}
}
/** All or nothing: when it fails, the cache stays as it was. */
public function replaceAll(string $repositoryId, array $packs): void {
$this->transaction(function () use ($repositoryId, $packs): void {
$qb = $this->db->getQueryBuilder();
$qb->delete('ngb_index')->where($qb->expr()->eq('repository_id', $qb->createNamedParameter($repositoryId)));
$qb->executeStatement();
$this->insertAll($repositoryId, $packs);
});
}
/** Run $write in a transaction of its own, or as part of the caller's when there is one. */
private function transaction(\Closure $write): void {
if ($this->db->inTransaction()) {
$write();
return;
}
$this->db->beginTransaction();
try {
$write();
$this->db->commit();
} catch (\Throwable $e) {
if ($this->db->inTransaction()) {
$this->db->rollBack();
}
throw $e;
}
}
/** @param array<string, list<array{id:string, offset:int, length:int, raw:int, flags:int}>> $packs */
private function insertAll(string $repositoryId, array $packs): void {
foreach ($packs as $packId => $entries) {
$this->insert($repositoryId, $packId, $entries);
}
}
/** @param list<array{id:string, offset:int, length:int, raw:int, flags:int}> $entries */
private function insert(string $repositoryId, string $packId, array $entries): void {
$qb = $this->db->getQueryBuilder();
$qb->insert('ngb_index')->values([
'repository_id' => $qb->createNamedParameter($repositoryId),
'pack_id' => $qb->createNamedParameter($packId),
'entries' => $qb->createNamedParameter(json_encode($entries, JSON_THROW_ON_ERROR)),
]);
try {
$qb->executeStatement();
} catch (Exception $e) {
if ($e->getReason() !== Exception::REASON_UNIQUE_CONSTRAINT_VIOLATION) {
throw $e;
}
}
$qb->executeStatement();
}
public function remove(string $repositoryId, array $packIds): void {