Hi, I would like to know exactly what this metric
...
# questions-and-troubleshootingu
이세찬
10/22/2025, 6:00 AMHi, I would like to know exactly what this metric
lake_get_tablet_metadata_countmeans.
k
Kevin Cai
10/22/2025, 6:05 AMthis is a counter for shared-data native table retrieving tablet metadata from remote storage.
u
이세찬
10/22/2025, 8:05 AM@Kevin Cai
This metric is increasing linearly. This means:
1. Metadata lookups on remote storage are steadily increasing.
2. If it's increasing linearly, could metadata lookups become increasingly slower? For example, could the response from show proc '/dbs' become slow, or could the stream load response fail?
3. Is there an upper limit for this metric?
Can you elaborate or point me to documentation?
이세찬
10/22/2025, 8:07 AM Copy code
lake_get_tablet_metadata_count{namespace="~~~"}k
Kevin Cai
10/22/2025, 8:18 AMits an accumulated counter, so the number will increase infinitely, shall check the rate(lake_get_tablet_metadata_count), if there is an increment.
✅ 1
u
이세찬
10/22/2025, 8:18 AMIs there an upper limit for this metric?
이세찬
10/22/2025, 8:21 AMlake_get_tablet_metadata_latency_99
this metric also increase infinitely ?
k
Kevin Cai
10/22/2025, 8:24 AMIs there an upper limit for this metric?no, it is a int64 counter.
✅ 1
u
이세찬
10/22/2025, 8:26 AM Copy code
lake_get_tablet_metadata_latency_9999
lake_get_tablet_metadata_latency_99
lake_get_tablet_metadata_latency_90
lake_get_tablet_metadata_latency_80
lake_get_tablet_metadata_max_latency Copy code
Is there a description of these metrics?
I can't find it in the official Starrocks documentation, so I'm asking.k
Kevin Cai
10/22/2025, 8:28 AMcan refer to this doc: https://docs.starrocks.io/docs/administration/management/monitoring/Monitor_and_Alert/#125-configure-dashboard
It is included in the "Dashboard template for Shared-data Cluster - General" dashboard.
Kevin Cai
10/22/2025, 8:29 AMu
이세찬
10/23/2025, 6:39 AM@Kevin Cai I have a question.
I upgraded from version 3.4.5 to 3.5.6.
This is a shared-data deployment.
After the upgrade, stream load kept failing.
The symptoms were as follows:
1. I couldn’t access the information_schema tables.
2.
SHOW PROC '/dbs' Copy code
starrocks-fe-2 fe 2025-10-21 21:58:59.763+09:00 WARN (tablet stat mgr|30)
[TabletStatMgr$CollectTabletStatJob.sendTasks():382] Skip sending tablet stat task
for partition db_name.table_name.948008 because exception:
No alive backend or compute node in warehouse name: default_warehouse.이세찬
10/23/2025, 6:41 AMThis symptom persisted, so I downgraded to 3.4.5. Even after downgrading, the problem persisted until stream load stopped and statements like "show proc '/dbs'" were executed.
k
Kevin Cai
10/23/2025, 6:47 AMwhat's the result of 'show compute nodes' or 'show backends' when the issue happens.
u
이세찬
10/23/2025, 7:20 AMwhen the issue happens
> always all compute node is alive
이세찬
10/23/2025, 8:42 AMEven now, I couldn't access the information_schema tables.
SHOW PROC '/dbs' didn't work either.
I'm experiencing two symptoms.
I'm using a spot instance for the compute node.