github-actions[bot] commented on code in PR #70791:
URL: https://github.com/apache/doris/pull/70791#discussion_r4639755292


##########
be/src/cloud/config.cpp:
##########
@@ -88,8 +88,8 @@ DEFINE_mInt32(delete_bitmap_lock_expiration_seconds, "10");
 DEFINE_mInt32(get_delete_bitmap_lock_max_retry_times, "100");
 
 DEFINE_mBool(enable_sync_tablet_delete_bitmap_by_cache, "true");
-DEFINE_mInt32(delete_bitmap_store_write_version, "1");
-DEFINE_mInt32(delete_bitmap_store_read_version, "1");
+DEFINE_mInt32(delete_bitmap_store_write_version, "2");

Review Comment:
   [P1] Make the v2-only write default rollout-safe. After this default is 
picked up, an upgraded BE sends only v2 delete-bitmaps, while an older BE still 
using its v1-only default asks the Meta Service for only v1 and receives 
success without those entries. If that older BE serves a query or synchronizes 
the rowset, superseded rows remain visible. The local check at 
`cloud_meta_mgr.cpp:1202-1210` cannot detect another BE's reader mode. Keep 
dual writes (version 3) or enforce an instance-wide dual-read gate before any 
v2-only writer, and document/test the required order.



##########
be/src/cloud/config.cpp:
##########
@@ -88,8 +88,8 @@ DEFINE_mInt32(delete_bitmap_lock_expiration_seconds, "10");
 DEFINE_mInt32(get_delete_bitmap_lock_max_retry_times, "100");
 
 DEFINE_mBool(enable_sync_tablet_delete_bitmap_by_cache, "true");
-DEFINE_mInt32(delete_bitmap_store_write_version, "1");
-DEFINE_mInt32(delete_bitmap_store_read_version, "1");
+DEFINE_mInt32(delete_bitmap_store_write_version, "2");
+DEFINE_mInt32(delete_bitmap_store_read_version, "3");

Review Comment:
   [P2] Prevent the dual-read default from rejecting large existing bitmaps. 
`store_version=3` sums the v1 and v2 payloads into `delete_bitmap_byte` before 
the BE merges them; a dual-written rowset can cross the 1 GiB 
`max_get_delete_bitmap_byte` cap even though each single-format read is below 
it, causing `KV_TXN_GET_ERR` and failed tablet synchronization. Account for 
equivalent dual copies separately or provide batching/limit handling for one 
rowset's dual-format payload.



##########
be/src/cloud/config.cpp:
##########
@@ -88,8 +88,8 @@ DEFINE_mInt32(delete_bitmap_lock_expiration_seconds, "10");
 DEFINE_mInt32(get_delete_bitmap_lock_max_retry_times, "100");
 
 DEFINE_mBool(enable_sync_tablet_delete_bitmap_by_cache, "true");
-DEFINE_mInt32(delete_bitmap_store_write_version, "1");
-DEFINE_mInt32(delete_bitmap_store_read_version, "1");
+DEFINE_mInt32(delete_bitmap_store_write_version, "2");

Review Comment:
   [P1] Gate this default on a v2-capable Meta Service. A pre-v2 Meta Service 
ignores the new v2 request fields, falls back to `store_version=1`, sees no v1 
rows, commits an empty update, and returns OK; the new BE therefore silently 
loses the delete bitmap. Upgrade the Meta Service first with a capability 
check, or retain a compatible dual-write path until that upgrade is complete.



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to