Vladimir Steshin created IGNITE-20507:
-----------------------------------------
Summary: Persistent cache meta is not removed if node filter skips
node.
Key: IGNITE-20507
URL: https://issues.apache.org/jira/browse/IGNITE-20507
Project: Ignite
Issue Type: Bug
Reporter: Vladimir Steshin
We keep persistent cache meta on node, which is filtered by the cache node
filter. If such cache is removed, some nodes can retain 'cache_data.dat'.
Such nodes can't re-join cluster because they find this 'cache_data.dat' and
offer this cache when joining the cluster. But the cache has been removed:
{code:java}
org.apache.ignite.spi.IgniteSpiException: Joining node has caches with data
which are not presented on cluster, it could mean that they were already
destroyed, to add the node to cluster - remove directories with the
caches[TestDynamicCache]
{code}
This happens because we remove persistent cache data in
`GridCacheProcessor#prepareCacheStop` looking at `Map<String,
GridCacheAdapter<?, ?>> GridCacheProcessor#caches`. But there is no
GridCacheAdapter for the cache if the node filter excludes this cache for
thecurrent node. But 'cache_data.dat' exists.
The work-around is to acquire a cache proxy on the node for which this cache is
excluded:
{code:java}
// This fixes the issue!
// cache = grid(2).cache(cfg.getName());
{code}
This creates the proxy and registers missing GridCacheAdapter.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
