[
https://issues.apache.org/jira/browse/IGNITE-586?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14790906#comment-14790906
]
Ivan Veselovsky commented on IGNITE-586:
----------------------------------------
We revealed that the primary cause of the problem is that some tests bring IGFS
into incorrect state: some entries in meta & data caches exist, but cannot be
found by treversing the file system tree.
The fact is that at least renames cause such situation. The problem is
localized tightly enough.
This happens at least in situation when a file cannot be moved to a destination
folder where a file with the same short name already exists: in such situation
the source file "disappears" from the file system, while still persists in meta
cache.
> [Test] GridGgfsAbstractSelfTest.testFormat() fails on TC.
> ---------------------------------------------------------
>
> Key: IGNITE-586
> URL: https://issues.apache.org/jira/browse/IGNITE-586
> Project: Ignite
> Issue Type: Bug
> Components: hadoop
> Reporter: Artem Shutak
> Assignee: Ivan Veselovsky
> Labels: failedtest
> Fix For: ignite-1.5
>
>
> The problem is that format() doesn't clean all previously written data.
> See GG-8578 (the issue was closed, but stay reproducible)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
