[ https://issues.apache.org/jira/browse/SOLR-14061?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16994157#comment-16994157 ]
Robert Muir commented on SOLR-14061: ------------------------------------ It seems like maybe solr tries to replicate this corruption too: {{ [junit4] 2> 382777 ERROR (indexFetcher-1233-thread-1) [ ] o.a.s.h.IndexFetcher Error fetching file, doing one retry...:org.apache.solr.common.SolrException: Unable to download _2_Asserting_0.dvd completely. Downloaded 0!=85}} Again, I don't want to raise false alarms, but we should make these kinds of things impossible. > TestLocalFSCloudBackupRestore may have exposed corruption possibility > --------------------------------------------------------------------- > > Key: SOLR-14061 > URL: https://issues.apache.org/jira/browse/SOLR-14061 > Project: Solr > Issue Type: Bug > Security Level: Public(Default Security Level. Issues are Public) > Reporter: Robert Muir > Priority: Major > Attachments: SOLR-14061.log > > > This test really expects "/location/does/not/exist" to not exist. It seems to > pass (most of the time). But running on my computer under heavy load, it > fails with AlreadyClosedException from CachingDirectoryFactory, > NoSuchFileException from lucene for index files, threads and resources > leaking everywhere, it dies pretty violently. > And I really don't like that I see NoSuchFileException for files in the > underlying lucene index files in the logs: this seems really bad: looks like > corruption to me. > So I think some exceptions or errors (SecurityException in this case) might > be unexpected and cause something bad to happen. This could happen in > production with something like OutOfMemoryError or IOError that is rare but > currently handled incorrectly. > I will open a separate issue to grant read access to > "/location/does/not/exist" so that the test works as designed, but I want to > file this one in case there is the chance that an unexpected Throwable of > some kind can cause corruption. I will upload logs and seeds and stuff. > please close as won't fix if this is a false alarm, I didn't review the code, > I just read the damage. -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org