[ 
https://issues.apache.org/jira/browse/IGNITE-9854?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16660516#comment-16660516
 ] 

Nikolay Izhikov commented on IGNITE-9854:
-----------------------------------------

Hello, [~ivandasch], [~DmitriyGovorukhin].

Can you ensure patch is complete?
>From what I see:
{code:java}
@Override public GridMultiCollectionWrapper<FullPageId> beginCheckpoint() 
throws IgniteException {
    //...
    collections[i] = seg.segCheckpointPages = seg.dirtyPages;

    seg.dirtyPages = new GridConcurrentHashSet<>();
    //...
}
{code}
And your code:
{code:java}
    private long refreshOutdatedPage(Segment seg, int grpId, long pageId, 
boolean rmv) {
        //...
        Collection<FullPageId> cpPages = seg.segCheckpointPages;

        if (cpPages != null)
            cpPages.remove(new FullPageId(pageId, grpId));

        Collection<FullPageId> dirtyPages = seg.dirtyPages;

        if (dirtyPages != null)
            dirtyPages.remove(new FullPageId(pageId, grpId));
        //...
    }
{code}
As far as I can understand there is a specific execution path when we don't see 
`segCheckpointPages` nor `dirtyPages`
{noformat}
T1 - Collection<FullPageId> cpPages = seg.segCheckpointPages;
T2 - seg.segCheckpointPages = seg.dirtyPages;
T2 - seg.dirtyPages = new GridConcurrentHashSet<>();
T1 - dirtyPages.remove(new FullPageId(pageId, grpId));
{noformat}

> NullPointerException in PageMemoryImpl.refreshOutdatedPages during removing 
> from segCheckpointPages
> ---------------------------------------------------------------------------------------------------
>
>                 Key: IGNITE-9854
>                 URL: https://issues.apache.org/jira/browse/IGNITE-9854
>             Project: Ignite
>          Issue Type: Bug
>          Components: persistence
>    Affects Versions: 2.6
>            Reporter: Ivan Daschinskiy
>            Assignee: Ivan Daschinskiy
>            Priority: Critical
>             Fix For: 2.8
>
>
> Because of possibility of concurrently setting segCheckpointPages to null of 
> segment not under segment writeLock (i.e. in 
> PageMemoryImpl#finishCheckpoint), NullPointerException is possible. This 
> causes immediate node failure. 
> Example stack trace is attached (failure during iteration in rebalance 
> supplier).
> {code:java}
> java.lang.NullPointerException: null
>         at 
> org.apache.ignite.internal.processors.cache.persistence.pagemem.PageMemoryImpl.refreshOutdatedPage(PageMemoryImpl.java:840)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.pagemem.PageMemoryImpl.access$5100(PageMemoryImpl.java:120)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.pagemem.PageMemoryImpl$Segment.removePageForReplacement(PageMemoryImpl.java:2175)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.pagemem.PageMemoryImpl$Segment.access$900(PageMemoryImpl.java:1841)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.pagemem.PageMemoryImpl.acquirePage(PageMemoryImpl.java:686)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.pagemem.PageMemoryImpl.acquirePage(PageMemoryImpl.java:627)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.CacheDataRowAdapter.initFromLink(CacheDataRowAdapter.java:140)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.CacheDataRowAdapter.initFromLink(CacheDataRowAdapter.java:102)
>         at 
> org.apache.ignite.internal.processors.cache.tree.DataRow.<init>(DataRow.java:54)
>         at 
> org.apache.ignite.internal.processors.cache.tree.CacheDataRowStore.dataRow(CacheDataRowStore.java:73)
>         at 
> org.apache.ignite.internal.processors.cache.tree.CacheDataTree.getRow(CacheDataTree.java:146)
>         at 
> org.apache.ignite.internal.processors.cache.tree.CacheDataTree.getRow(CacheDataTree.java:41)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.tree.BPlusTree$ForwardCursor.fillFromBuffer(BPlusTree.java:4660)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.tree.BPlusTree$ForwardCursor.nextPage(BPlusTree.java:4760)
>         at 
> org.apache.ignite.internal.processors.cache.persistence.tree.BPlusTree$ForwardCursor.next(BPlusTree.java:4689)
> {code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to