/** returns dynamic Permissions to configured paths */
static void addFilePermissions(Permissions policy, Environment environment) throws IOException {
// read-only dirs
addPath(policy, "path.home", environment.binFile(), "read,readlink");
addPath(policy, "path.home", environment.libFile(), "read,readlink");
addPath(policy, "path.plugins", environment.pluginsFile(), "read,readlink");
addPath(policy, "path.conf", environment.configFile(), "read,readlink");
addPath(policy, "path.scripts", environment.scriptsFile(), "read,readlink");
// read-write dirs
addPath(policy, "java.io.tmpdir", environment.tmpFile(), "read,readlink,write,delete");
addPath(policy, "path.logs", environment.logsFile(), "read,readlink,write,delete");
if (environment.sharedDataFile() != null) {
addPath(
policy, "path.shared_data", environment.sharedDataFile(), "read,readlink,write,delete");
}
for (Path path : environment.dataFiles()) {
addPath(policy, "path.data", path, "read,readlink,write,delete");
}
for (Path path : environment.dataWithClusterFiles()) {
addPath(policy, "path.data", path, "read,readlink,write,delete");
}
for (Path path : environment.repoFiles()) {
addPath(policy, "path.repo", path, "read,readlink,write,delete");
}
if (environment.pidFile() != null) {
// we just need permission to remove the file if its elsewhere.
policy.add(new FilePermission(environment.pidFile().toString(), "delete"));
}
}
/** Adds access to all configurable paths. */
static void addFilePermissions(Permissions policy, Environment environment) {
// read-only dirs
addPath(policy, Environment.PATH_HOME_SETTING.getKey(), environment.binFile(), "read,readlink");
addPath(policy, Environment.PATH_HOME_SETTING.getKey(), environment.libFile(), "read,readlink");
addPath(
policy, Environment.PATH_HOME_SETTING.getKey(), environment.modulesFile(), "read,readlink");
addPath(
policy, Environment.PATH_HOME_SETTING.getKey(), environment.pluginsFile(), "read,readlink");
addPath(
policy, Environment.PATH_CONF_SETTING.getKey(), environment.configFile(), "read,readlink");
addPath(
policy,
Environment.PATH_SCRIPTS_SETTING.getKey(),
environment.scriptsFile(),
"read,readlink");
// read-write dirs
addPath(policy, "java.io.tmpdir", environment.tmpFile(), "read,readlink,write,delete");
addPath(
policy,
Environment.PATH_LOGS_SETTING.getKey(),
environment.logsFile(),
"read,readlink,write,delete");
if (environment.sharedDataFile() != null) {
addPath(
policy,
Environment.PATH_SHARED_DATA_SETTING.getKey(),
environment.sharedDataFile(),
"read,readlink,write,delete");
}
for (Path path : environment.dataFiles()) {
addPath(policy, Environment.PATH_DATA_SETTING.getKey(), path, "read,readlink,write,delete");
}
// TODO: this should be removed in ES 6.0! We will no longer support data paths with the cluster
// as a folder
assert Version.CURRENT.major < 6 : "cluster name is no longer used in data path";
for (Path path : environment.dataWithClusterFiles()) {
addPathIfExists(
policy, Environment.PATH_DATA_SETTING.getKey(), path, "read,readlink,write,delete");
}
for (Path path : environment.repoFiles()) {
addPath(policy, Environment.PATH_REPO_SETTING.getKey(), path, "read,readlink,write,delete");
}
if (environment.pidFile() != null) {
// we just need permission to remove the file if its elsewhere.
policy.add(new FilePermission(environment.pidFile().toString(), "delete"));
}
}
private void testLoad(boolean fullRecovery) {
logger.info("Running with fullRecover [{}]", fullRecovery);
startNode("server1");
logger.info("Running Cluster Health (waiting for node to startup properly)");
ClusterHealthResponse clusterHealth =
client("server1")
.admin()
.cluster()
.health(clusterHealthRequest().waitForGreenStatus())
.actionGet();
logger.info("Done Cluster Health, status " + clusterHealth.getStatus());
assertThat(clusterHealth.isTimedOut(), equalTo(false));
assertThat(clusterHealth.getStatus(), equalTo(ClusterHealthStatus.GREEN));
// get the environment, so we can clear the work dir when needed
Environment environment =
((InternalNode) node("server1")).injector().getInstance(Environment.class);
logger.info("--> creating test index ...");
client("server1").admin().indices().prepareCreate("test").execute().actionGet();
logger.info("Running Cluster Health (wait for the shards to startup)");
clusterHealth =
client("server1")
.admin()
.cluster()
.health(clusterHealthRequest().waitForYellowStatus().waitForActiveShards(1))
.actionGet();
logger.info("Done Cluster Health, status " + clusterHealth.getStatus());
assertThat(clusterHealth.isTimedOut(), equalTo(false));
assertThat(clusterHealth.getStatus(), equalTo(ClusterHealthStatus.YELLOW));
logger.info("--> refreshing and checking count");
client("server1").admin().indices().prepareRefresh().execute().actionGet();
assertThat(
client("server1").prepareCount().setQuery(matchAllQuery()).execute().actionGet().getCount(),
equalTo(0l));
logger.info("--> indexing 1234 docs");
for (long i = 0; i < 1234; i++) {
client("server1")
.prepareIndex("test", "type1", Long.toString(i))
.setCreate(
true) // make sure we use create, so if we recover wrongly, we will get increments...
.setSource(MapBuilder.<String, Object>newMapBuilder().put("test", "value" + i).map())
.execute()
.actionGet();
// snapshot every 100 so we get some actions going on in the gateway
if ((i % 11) == 0) {
client("server1").admin().indices().prepareGatewaySnapshot().execute().actionGet();
}
// flush every once is a while, so we get different data
if ((i % 55) == 0) {
client("server1").admin().indices().prepareFlush().execute().actionGet();
}
}
logger.info("--> refreshing and checking count");
client("server1").admin().indices().prepareRefresh().execute().actionGet();
assertThat(
client("server1").prepareCount().setQuery(matchAllQuery()).execute().actionGet().getCount(),
equalTo(1234l));
logger.info("--> closing the server");
closeNode("server1");
if (fullRecovery) {
logger.info("Clearing cluster data dir, so there will be a full recovery from the gateway");
FileSystemUtils.deleteRecursively(environment.dataWithClusterFiles());
logger.info(
"Starting the server, should recover from the gateway (both index and translog) without reusing work dir");
}
startNode("server1");
logger.info("--> running Cluster Health (wait for the shards to startup)");
clusterHealth =
client("server1")
.admin()
.cluster()
.health(clusterHealthRequest().waitForYellowStatus().waitForActiveShards(1))
.actionGet();
logger.info("--> done Cluster Health, status " + clusterHealth.getStatus());
assertThat(clusterHealth.isTimedOut(), equalTo(false));
assertThat(clusterHealth.getStatus(), equalTo(ClusterHealthStatus.YELLOW));
logger.info("--> checking count");
assertThat(
client("server1").prepareCount().setQuery(matchAllQuery()).execute().actionGet().getCount(),
equalTo(1234l));
logger.info("--> checking reuse / recovery status");
IndicesStatusResponse statusResponse =
client("server1").admin().indices().prepareStatus().setRecovery(true).execute().actionGet();
for (IndexShardStatus indexShardStatus : statusResponse.getIndex("test")) {
for (ShardStatus shardStatus : indexShardStatus) {
if (shardStatus.getShardRouting().primary()) {
if (fullRecovery || !isPersistentStorage()) {
assertThat(
shardStatus.getGatewayRecoveryStatus().getReusedIndexSize().bytes(), equalTo(0l));
} else {
assertThat(
shardStatus.getGatewayRecoveryStatus().getReusedIndexSize().bytes(),
greaterThan(
shardStatus.getGatewayRecoveryStatus().getIndexSize().bytes()
- 8196 /* segments file and others */));
}
}
}
}
}
@Test
@Slow
public void testSnapshotOperations() throws Exception {
startNode("server1", getClassDefaultSettings());
// get the environment, so we can clear the work dir when needed
Environment environment =
((InternalNode) node("server1")).injector().getInstance(Environment.class);
logger.info("Running Cluster Health (waiting for node to startup properly)");
ClusterHealthResponse clusterHealth =
client("server1")
.admin()
.cluster()
.health(clusterHealthRequest().waitForGreenStatus())
.actionGet();
logger.info("Done Cluster Health, status " + clusterHealth.getStatus());
assertThat(clusterHealth.isTimedOut(), equalTo(false));
assertThat(clusterHealth.getStatus(), equalTo(ClusterHealthStatus.GREEN));
// Translog tests
logger.info("Creating index [{}]", "test");
client("server1").admin().indices().prepareCreate("test").execute().actionGet();
// create a mapping
PutMappingResponse putMappingResponse =
client("server1")
.admin()
.indices()
.preparePutMapping("test")
.setType("type1")
.setSource(mappingSource())
.execute()
.actionGet();
assertThat(putMappingResponse.isAcknowledged(), equalTo(true));
// verify that mapping is there
ClusterStateResponse clusterState =
client("server1").admin().cluster().state(clusterStateRequest()).actionGet();
assertThat(clusterState.getState().metaData().index("test").mapping("type1"), notNullValue());
// create two and delete the first
logger.info("Indexing #1");
client("server1")
.index(Requests.indexRequest("test").type("type1").id("1").source(source("1", "test")))
.actionGet();
logger.info("Indexing #2");
client("server1")
.index(Requests.indexRequest("test").type("type1").id("2").source(source("2", "test")))
.actionGet();
// perform snapshot to the index
logger.info("Gateway Snapshot");
client("server1").admin().indices().gatewaySnapshot(gatewaySnapshotRequest("test")).actionGet();
logger.info("Deleting #1");
client("server1").delete(deleteRequest("test").type("type1").id("1")).actionGet();
// perform snapshot to the index
logger.info("Gateway Snapshot");
client("server1").admin().indices().gatewaySnapshot(gatewaySnapshotRequest("test")).actionGet();
logger.info("Gateway Snapshot (should be a no op)");
// do it again, it should be a no op
client("server1").admin().indices().gatewaySnapshot(gatewaySnapshotRequest("test")).actionGet();
logger.info("Closing the server");
closeNode("server1");
logger.info(
"Starting the server, should recover from the gateway (only translog should be populated)");
startNode("server1");
logger.info("Running Cluster Health (wait for the shards to startup)");
clusterHealth =
client("server1")
.admin()
.cluster()
.health(clusterHealthRequest().waitForYellowStatus().waitForActiveShards(1))
.actionGet();
logger.info("Done Cluster Health, status " + clusterHealth.getStatus());
assertThat(clusterHealth.isTimedOut(), equalTo(false));
assertThat(clusterHealth.getStatus(), equalTo(ClusterHealthStatus.YELLOW));
// verify that mapping is there
clusterState = client("server1").admin().cluster().state(clusterStateRequest()).actionGet();
assertThat(clusterState.getState().metaData().index("test").mapping("type1"), notNullValue());
logger.info("Getting #1, should not exists");
GetResponse getResponse =
client("server1").get(getRequest("test").type("type1").id("1")).actionGet();
assertThat(getResponse.isExists(), equalTo(false));
logger.info("Getting #2");
getResponse = client("server1").get(getRequest("test").type("type1").id("2")).actionGet();
assertThat(getResponse.getSourceAsString(), equalTo(source("2", "test")));
// Now flush and add some data (so we have index recovery as well)
logger.info(
"Flushing, so we have actual content in the index files (#2 should be in the index)");
client("server1").admin().indices().flush(flushRequest("test")).actionGet();
logger.info("Indexing #3, so we have something in the translog as well");
client("server1")
.index(Requests.indexRequest("test").type("type1").id("3").source(source("3", "test")))
.actionGet();
logger.info("Gateway Snapshot");
client("server1").admin().indices().gatewaySnapshot(gatewaySnapshotRequest("test")).actionGet();
logger.info("Gateway Snapshot (should be a no op)");
client("server1").admin().indices().gatewaySnapshot(gatewaySnapshotRequest("test")).actionGet();
logger.info("Closing the server");
closeNode("server1");
logger.info(
"Starting the server, should recover from the gateway (both index and translog) and reuse work dir");
startNode("server1");
logger.info("Running Cluster Health (wait for the shards to startup)");
clusterHealth =
client("server1")
.admin()
.cluster()
.health(clusterHealthRequest().waitForYellowStatus().waitForActiveShards(1))
.actionGet();
logger.info("Done Cluster Health, status " + clusterHealth.getStatus());
assertThat(clusterHealth.isTimedOut(), equalTo(false));
assertThat(clusterHealth.getStatus(), equalTo(ClusterHealthStatus.YELLOW));
logger.info("Getting #1, should not exists");
getResponse = client("server1").get(getRequest("test").type("type1").id("1")).actionGet();
assertThat(getResponse.isExists(), equalTo(false));
logger.info("Getting #2 (not from the translog, but from the index)");
getResponse = client("server1").get(getRequest("test").type("type1").id("2")).actionGet();
assertThat(getResponse.getSourceAsString(), equalTo(source("2", "test")));
logger.info("Getting #3 (from the translog)");
getResponse = client("server1").get(getRequest("test").type("type1").id("3")).actionGet();
assertThat(getResponse.getSourceAsString(), equalTo(source("3", "test")));
logger.info("Closing the server");
closeNode("server1");
logger.info("Clearing cluster data dir, so there will be a full recovery from the gateway");
FileSystemUtils.deleteRecursively(environment.dataWithClusterFiles());
logger.info(
"Starting the server, should recover from the gateway (both index and translog) without reusing work dir");
startNode("server1");
logger.info("Running Cluster Health (wait for the shards to startup)");
clusterHealth =
client("server1")
.admin()
.cluster()
.health(clusterHealthRequest().waitForYellowStatus().waitForActiveShards(1))
.actionGet();
logger.info("Done Cluster Health, status " + clusterHealth.getStatus());
assertThat(clusterHealth.isTimedOut(), equalTo(false));
assertThat(clusterHealth.getStatus(), equalTo(ClusterHealthStatus.YELLOW));
logger.info("Getting #1, should not exists");
getResponse = client("server1").get(getRequest("test").type("type1").id("1")).actionGet();
assertThat(getResponse.isExists(), equalTo(false));
logger.info("Getting #2 (not from the translog, but from the index)");
getResponse = client("server1").get(getRequest("test").type("type1").id("2")).actionGet();
assertThat(getResponse.getSourceAsString(), equalTo(source("2", "test")));
logger.info("Getting #3 (from the translog)");
getResponse = client("server1").get(getRequest("test").type("type1").id("3")).actionGet();
assertThat(getResponse.getSourceAsString(), equalTo(source("3", "test")));
logger.info(
"Flushing, so we have actual content in the index files (#3 should be in the index now as well)");
client("server1").admin().indices().flush(flushRequest("test")).actionGet();
logger.info("Gateway Snapshot");
client("server1").admin().indices().gatewaySnapshot(gatewaySnapshotRequest("test")).actionGet();
logger.info("Gateway Snapshot (should be a no op)");
client("server1").admin().indices().gatewaySnapshot(gatewaySnapshotRequest("test")).actionGet();
logger.info("Closing the server");
closeNode("server1");
logger.info(
"Starting the server, should recover from the gateway (just from the index, nothing in the translog)");
startNode("server1");
logger.info("Running Cluster Health (wait for the shards to startup)");
clusterHealth =
client("server1")
.admin()
.cluster()
.health(clusterHealthRequest().waitForYellowStatus().waitForActiveShards(1))
.actionGet();
logger.info("Done Cluster Health, status " + clusterHealth.getStatus());
assertThat(clusterHealth.isTimedOut(), equalTo(false));
assertThat(clusterHealth.getStatus(), equalTo(ClusterHealthStatus.YELLOW));
logger.info("Getting #1, should not exists");
getResponse = client("server1").get(getRequest("test").type("type1").id("1")).actionGet();
assertThat(getResponse.isExists(), equalTo(false));
logger.info("Getting #2 (not from the translog, but from the index)");
getResponse = client("server1").get(getRequest("test").type("type1").id("2")).actionGet();
assertThat(getResponse.getSourceAsString(), equalTo(source("2", "test")));
logger.info("Getting #3 (not from the translog, but from the index)");
getResponse = client("server1").get(getRequest("test").type("type1").id("3")).actionGet();
assertThat(getResponse.getSourceAsString(), equalTo(source("3", "test")));
logger.info("Deleting the index");
client("server1").admin().indices().delete(deleteIndexRequest("test")).actionGet();
}