[SENTRY-1759] UpdatableCache leaks connections

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.8.0, 2.0.0
• Fix Version/s: 1.8.0, 2.0.0
• Component/s: Sentry
• Assignee: Alex Kolbasov
• Reporter: Alex Kolbasov
• Created: 10/May/17 06:10
• Updated: 24/Jul/17 14:51
• Resolved: 10/May/17 16:44

Description

In UpdatableCache.loadFromRemote():

  private Table<String, String, Set<String>> loadFromRemote() throws Exception {
    Table<String, String, Set<String>> tempCache = HashBasedTable.create();
    String requestor;
    requestor = UserGroupInformation.getLoginUser().getShortUserName();

    final SentryGenericServiceClient client = getClient(); // <-- Client created
    final Set<TSentryRole> tSentryRoles = client.listAllRoles(requestor, componentType);

    for (TSentryRole tSentryRole : tSentryRoles) {
      final String roleName = tSentryRole.getRoleName();
      final Set<TSentryPrivilege> tSentryPrivileges = client.listPrivilegesByRoleName(requestor, roleName, componentType, serviceName);
      for (String group : tSentryRole.getGroups()) {
        Set<String> currentPrivileges = tempCache.get(group, roleName);
        if (currentPrivileges == null) {
          currentPrivileges = new HashSet<>();
          tempCache.put(group, roleName, currentPrivileges);
        }
        for (TSentryPrivilege tSentryPrivilege : tSentryPrivileges) {
          currentPrivileges.add(tSentryPrivilegeConverter.toString(tSentryPrivilege));
        }
      }
    }
    return tempCache;
  }

As you can see, the client is never closed, so on each timer iteration we are leaking one connected client. This may quickly exhaust server connection queue.

Issue Links

• blocks: SENTRY-872 Uber jira for HMS HA + Sentry HA redesign
• links to: Code Review