[
https://issues.apache.org/jira/browse/IGNITE-22844?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Maksim Davydov updated IGNITE-22844:
------------------------------------
Description:
Currently 'Read Repair' task performed for the chosen partitions by control.sh
requires CACHE_READ and CACHE PUT permissions in addition to ADMIN_OPS. Thus,
to perform the comand `control.sh --consistency repair` the one needs all 3
permissions.
There is no point to have additional CACHE_READ and CACHE_PUT permissions to
perform the operation, as they are useless for the task itself, and introduces
security risk, by allowing the user to manipulate the cache in parallel.
The solution would be to substitute user's security context with the context of
the cluster node that performs the command. The former is used for the
ADMIN_OPS permission check and the latter is used for cache. This will ease the
requirement for the users to perform `control.sh --consistency repair` without
introducing any additional security risk.
was:
Currently 'Read Repair' task performed for the chosen partitions by control.sh
requires additional CACHE_READ and CACHE PUT permissions in addition to
ADMIN_OPS.
This is not convinient, as it obliges users to have this additional
permissions, when it's not always neccesary.
> control.sh --consistency repair permissions fixes
> -------------------------------------------------
>
> Key: IGNITE-22844
> URL: https://issues.apache.org/jira/browse/IGNITE-22844
> Project: Ignite
> Issue Type: Improvement
> Reporter: Maksim Davydov
> Assignee: Maksim Davydov
> Priority: Minor
> Labels: ise
> Fix For: 2.17
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Currently 'Read Repair' task performed for the chosen partitions by
> control.sh requires CACHE_READ and CACHE PUT permissions in addition to
> ADMIN_OPS. Thus, to perform the comand `control.sh --consistency repair` the
> one needs all 3 permissions.
> There is no point to have additional CACHE_READ and CACHE_PUT permissions to
> perform the operation, as they are useless for the task itself, and
> introduces security risk, by allowing the user to manipulate the cache in
> parallel.
> The solution would be to substitute user's security context with the context
> of the cluster node that performs the command. The former is used for the
> ADMIN_OPS permission check and the latter is used for cache. This will ease
> the requirement for the users to perform `control.sh --consistency repair`
> without introducing any additional security risk.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
