[CASSANDRA-9953] Snapshot file handlers are not released after snapshot deleted

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Duplicate
• Severity: Normal
• Assignee: Unassigned
• Reporter: Imri Zvik
• Created: 02/Aug/15 08:00
• Updated: 16/Apr/19 09:31
• Resolved: 06/Aug/15 15:52

Description

We are seeing a lot of opened file descriptors to deleted snapshots (deleted using nodetool clearsnapshot):

java    128657 cassandra  DEL    REG              253,2              569272514 /var/lib/cassandra/data/accounts/account_store_data/snapshots/feb5f790-316e-11e5-aec0-472b0d6e3fd4/accounts-account_store_data-jb-264593-Index.db
java    128657 cassandra  DEL    REG              253,2             1610616657 /var/lib/cassandra/data/accounts/account_store_counters/snapshots/03aa4710-316f-11e5-aec0-472b0d6e3fd4/accounts-account_store_counters-jb-635527-Index.db
java    128657 cassandra  DEL    REG              253,2             1610613856 /var/lib/cassandra/data/accounts/account_store_counters/snapshots/43c17170-316f-11e5-aec0-472b0d6e3fd4/accounts-account_store_counters-jb-635675-Index.db
java    128657 cassandra  DEL    REG              253,2             1610613052 /var/lib/cassandra/data/accounts/account_store_counters/snapshots/18e001a0-3170-11e5-aec0-472b0d6e3fd4/accounts-account_store_counters-jb-636200-Index.db

[root@cassandra002 ~]# lsof -np 128657 |grep -c DEL
56682

They are probably created by the routine repair process, but they are never cleared (restarting the Cassandra process clears them, of course).
We are seeing these also after all repair processes finished, and no repair process is running in the cluster.

There are no errors or fatals in the system.log.
We are using Datastax community edition 2.0.13, installed from RPMs.