simges commented on issue #67934: URL: https://github.com/apache/doris/issues/67934#issuecomment-5740952879
> Hi, I've submitted a fix in [#67944](https://github.com/apache/doris/pull/67944) that aligns the FE default of `storage_flood_stage_usage_percent` to 90, consistent with the BE config. This prevents FE from scheduling load/restore jobs onto a BE whose disk is already in flood stage. Please take a look when you have time, thanks! @Asthenia0412 , many thanks for your kind response. But i believe this change may not be sufficient to fix this problem. Firstly, there is another condition to me met: ``` this.getUsedPct() > (Config.storage_flood_stage_usage_percent / 100.0) -> this condition can be addressed with your change diskAvailableCapacityB < Config.storage_flood_stage_left_capacity_bytes -> however, this will remain unresolved. ``` The proper solution, i believe, would be that BE nodes should signal FE back when their disk is full through Thrift API: 1. BE node signals FE besides regular DiskReports that it has no space left for new tablets. 2. FE gets this notification and discards that exact node when picking BE candidates suitable for query execution. Regards, -- 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]
