DAOS-XXXX cksum: retry once on checksum mismatch on update

Unlike fetch, we return DER_CSUM on update (turned into EIO by dfs) without any retry. We should retry at least once in case it is a transient error. The patch also prints more information about the actual checksum mismatch. Signed-off-by: Johann Lombardi <[email protected]>
daos-stack · Aug 27, 2024 · 4f48964 · 4f48964
1 parent e3869b7
commit 4f48964
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 3 deletions.
diff --git a/src/common/checksum.c b/src/common/checksum.c
@@ -280,6 +280,10 @@ daos_csummer_compare_csum_info(struct daos_csummer *obj,
 						  a->cs_len);
 	}
 
+	if (unlikely(!match))
+		D_ERROR("Checksum mismatch at index %d/%d "DF_CI" != "DF_CI"\n", i,  a->cs_nr,
+			DP_CI(ci_idx2csum(a, i)), DP_CI(ci_idx2csum(b, i)));
+
 	return match;
 }
 

diff --git a/src/object/cli_obj.c b/src/object/cli_obj.c
@@ -4760,9 +4760,17 @@ obj_comp_cb(tse_task_t *task, void *data)
 						obj_auxi->tx_uncertain = 1;
 					else
 						obj_auxi->nvme_io_err = 1;
-				} else if (task->dt_result != -DER_NVME_IO) {
-					/* Don't retry update for CSUM & UNCERTAIN errors */
-					obj_auxi->io_retry = 0;
+				} else {
+					if (task->dt_result == -DER_CSUM) {
+						/** Retry once on checksum error on update */
+						if (!obj_auxi->csum_retry)
+							obj_auxi->csum_retry = 1;
+						else
+							obj_auxi->io_retry = 0;
+					else if (task->dt_result != -DER_NVME_IO) {
+						/* Don't retry update for UNCERTAIN errors */
+						obj_auxi->io_retry = 0;
+					}
 				}
 			} else {
 				obj_auxi->io_retry = 0;