Old MV partitions not deleted in shared data mode from Minio

[swapkh91]

Steps to reproduce the behavior (Required)

CREATE TABLE `events` (
  `anonymousId` varchar(65533) NULL COMMENT "",
  `event` varchar(65533) NULL COMMENT "",
  `timestamp` datetime NULL COMMENT ""
) ENGINE=OLAP 
DUPLICATE KEY(`anonymousId`)
COMMENT "OLAP"
PARTITION BY date_trunc('DAY', timestamp)
DISTRIBUTED BY RANDOM
PROPERTIES (
"bloom_filter_columns" = "timestamp",
"bucket_size" = "4294967296",
"compression" = "LZ4",
"datacache.enable" = "true",
"datacache.partition_duration" = "1 days",
"enable_async_write_back" = "false",
"partition_live_number" = "1",
"replication_num" = "1",
"storage_volume" = "default_volume"
);

CREATE MATERIALIZED VIEW `events_mv` (`anonymousId`, `timestamp_ist`, `timestamp`,
  INDEX segment_id_idx (`anonymousId`) USING BITMAP COMMENT '')
COMMENT "MATERIALIZED_VIEW"
PARTITION BY (date_trunc('day', `timestamp`))
DISTRIBUTED BY RANDOM
REFRESH ASYNC START("2025-02-06 10:00:00") EVERY(INTERVAL 1 MINUTE)
PROPERTIES (
"replicated_storage" = "true",
"partition_ttl_number" = "2",
"replication_num" = "1",
"datacache.enable" = "true",
"enable_async_write_back" = "false",
"storage_volume" = "default_volume",
"warehouse" = "default_warehouse"
)
AS SELECT `events`.`timestamp_ist`, `events`.`timestamp`
FROM (SELECT convert_tz(`events`.`timestamp`, 'GMT', '+05:30') AS `timestamp_ist`, `events`.`timestamp`
FROM `segment`.`events`) `events`
WHERE (`events`.`event` = 'EXPIRED');

Expected behavior (Required)

show partitions from events_mv;
this command shows 1-2 partitions, but in Minio previous partition folders are still present. The older partition folders usually contain only SCHEMA file and sometimes meta folder and SCHEMA file both, refer screenshots.

Real behavior (Required)

This issue was present for tables too and resolved in #41675. Older folders should automatically be deleted from Minio.

StarRocks version (Required)

3.3.8

[srlch]

@swapkh91 Hi, is the cluster upgraded from v3.2? or other lower version?

[swapkh91]

@swapkh91 Hi, is the cluster upgraded from v3.2? or other lower version?

@srlch i upgraded it from 3.2.10 -> 3.3.1 -> 3.3.8

[srlch]

@swapkh91 before v3.3, the older partition folders can not be remove directly, could you reproduce this problem in 3.3.8 for a new MV?

[swapkh91]

@swapkh91 before v3.3, the older partition folders can not be remove directly, could you reproduce this problem in 3.3.8 for a new MV?

i'll check for new MV
for tables it was automatically resolved after upgrading, why not MV?

[srlch]

@swapkh91 before v3.3, the older partition folders can not be remove directly, could you reproduce this problem in 3.3.8 for a new MV?

i'll check for new MV for tables it was automatically resolved after upgrading, why not MV?

The partition removed after upgrading would be erased the folder, both for table and MV, i will also try to reproduce this problem.