Hi, Yesterday we've updatet our Graylog2/Elasticsearch Cluster. The Elastic Search Cluster consists of 3 physical Maschins: DL380 G7, E5620, 16GB RAM on RHEL 6.6. Each ES Node gets 4GB RAM. On one Host there is the graylog2 Server/Interface installed. Until yesterday we used Elasticsearch 0.90.10-1 and graylog2-0.20.3 Yesterday we updatet graylog2 to 0.90.0, startet everything, everything was running fine. Then Stopped graylog2 and the ElaticSearch Cluster, upgraded ES to 1.3.4 and graylog to 0.92.4. The Upgrade from ES was successfully, after that, startet graylog2, which connected to the cluster and showed everything.
In the ES Cluster there are 7 indices a 20mio messages. The last 3 indices are opened, the other closed. Graylog2 sees approx 50mio messages. New messages arrive with approx 5msg/sec In the logs from graylog2-server there are messages like this, every couple of minutes: org.graylog2.periodical.GarbageCollectionWarningThread - Last GC run with PS Scavenge took longer than 1 second It seems graylog is running fine, a bit slow on searches, but fine. Attached are the config files for graylog2 and elasticsearch. Can someone give us a hint where this warnings come from? What we can tweak? Would be very helpful! Thanks! Chris... -- You received this message because you are subscribed to the Google Groups "graylog2" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
elasticsearch.yml
Description: Binary data
# If you are running more than one instances of graylog2-server you have to select one of these # instances as master. The master will perform some periodical tasks that non-masters won't perform. is_master = true # The auto-generated node ID will be stored in this file and read after restarts. It is a good idea # to use an absolute file path here if you are starting graylog2-server from init scripts or similar. node_id_file = /etc/graylog2-server-node-id # You MUST set a secret to secure/pepper the stored user passwords here. Use at least 64 characters. # Generate one by using for example: pwgen -N 1 -s 96 password_secret = xxx # The default root user is named 'admin' #root_username = admin # You MUST specify a hash password for the root user (which you only need to initially set up the # system and in case you lose connectivity to your authentication backend) # This password cannot be changed using the API or via the web interface. If you need to change it, # modify it in this file. # Create one by using for example: echo -n yourpassword | shasum -a 256 # and put the resulting hash value into the following line root_password_sha2 = xxx # Set plugin directory here (relative or absolute) plugin_dir = plugin # REST API listen URI. Must be reachable by other graylog2-server nodes if you run a cluster. rest_listen_uri = http://127.0.0.1:12900/ # REST API transport address. Defaults to the value of rest_listen_uri. Exception: If rest_listen_uri # is set to a wildcard IP address (0.0.0.0) the first non-loopback IPv4 system address is used. # This will be promoted in the cluster discovery APIs and other nodes may try to connect on this # address. (see rest_listen_uri) #rest_transport_uri = http://192.168.1.1:12900/ # Enable CORS headers for REST API. This is necessary for JS-clients accessing the server directly. # If these are disabled, modern browsers will not be able to retrieve resources from the server. # This is disabled by default. Uncomment the next line to enable it. #rest_enable_cors = true # Enable GZIP support for REST API. This compresses API responses and therefore helps to reduce # overall round trip times. This is disabled by default. Uncomment the next line to enable it. #rest_enable_gzip = true # Enable HTTPS support for the REST API. This secures the communication with the REST API with # TLS to prevent request forgery and eavesdropping. This is disabled by default. Uncomment the # next line to enable it. #rest_enable_tls = true # The X.509 certificate file to use for securing the REST API. #rest_tls_cert_file = /path/to/graylog2.crt # The private key to use for securing the REST API. #rest_tls_key_file = /path/to/graylog2.key # The password to unlock the private key used for securing the REST API. #rest_tls_key_password = secret # The maximum size of a single HTTP chunk in bytes. #rest_max_chunk_size = 8192 # The maximum size of the HTTP request headers in bytes. #rest_max_header_size = 8192 # The maximal length of the initial HTTP/1.1 line in bytes. #rest_max_initial_line_length = 4096 # The size of the execution handler thread pool used exclusively for serving the REST API. #rest_thread_pool_size = 16 # The size of the worker thread pool used exclusively for serving the REST API. #rest_worker_threads_max_pool_size = 16 # Embedded Elasticsearch configuration file # pay attention to the working directory of the server, maybe use an absolute path here #elasticsearch_config_file = /etc/graylog2-elasticsearch.yml # Graylog2 will use multiple indices to store documents in. You can configured the strategy it uses to determine # when to rotate the currently active write index. # It supports multiple rotation strategies: # - "count" of messages per index, use elasticsearch_max_docs_per_index below to configure # - "size" per index, use elasticsearch_max_size_per_index below to configure # valid values are "count", "size" and "time", default is "count" rotation_strategy = count # (Approximate) maximum number of documents in an Elasticsearch index before a new index # is being created, also see no_retention and elasticsearch_max_number_of_indices. # Configure this if you used 'rotation_strategy = count' above. elasticsearch_max_docs_per_index = 20000000 # (Approximate) maximum size in bytes per Elasticsearch index on disk before a new index is being created, also see # no_retention and elasticsearch_max_number_of_indices. Default is 1GB. # Configure this if you used 'rotation_strategy = size' above. #elasticsearch_max_size_per_index = 1073741824 # (Approximate) maximum time before a new Elasticsearch index is being created, also see # no_retention and elasticsearch_max_number_of_indices. Default is 1 day. # Configure this if you used 'rotation_strategy = time' above. # Please note that this rotation period does not look at the time specified in the received messages, but is # using the real clock value to decide when to rotate the index! # Specify the time using a duration and a suffix indicating which unit you want: # 1w = 1 week # 1d = 1 day # 12h = 12 hours # Permitted suffixes are: d for day, h for hour, m for minute, s for second. #elasticsearch_max_time_per_index = 1d # Disable checking the version of Elasticsearch for being compatible with this Graylog2 release. # WARNING: Using Graylog2 with unsupported and untested versions of Elasticsearch may lead to data loss! #elasticsearch_disable_version_check = true # Disable message retention on this node, i. e. disable Elasticsearch index rotation. #no_retention = false # How many indices do you want to keep? # elasticsearch_max_number_of_indices*elasticsearch_max_docs_per_index=total number of messages in your setup elasticsearch_max_number_of_indices = 20 # Decide what happens with the oldest indices when the maximum number of indices is reached. # The following strategies are availble: # - delete # Deletes the index completely (Default) # - close # Closes the index and hides it from the system. Can be re-opened later. retention_strategy = close # How many Elasticsearch shards and replicas should be used per index? Note that this only applies to newly created indices. elasticsearch_shards = 3 elasticsearch_replicas = 2 # Prefix for all Elasticsearch indices and index aliases managed by Graylog2. elasticsearch_index_prefix = graylog2 # Do you want to allow searches with leading wildcards? This can be extremely resource hungry and should only # be enabled with care. See also: http://graylog2.org/resources/documentation/general/queries allow_leading_wildcard_searches = false # Do you want to allow searches to be highlighted? Depending on the size of your messages this can be memory hungry and # should only be enabled after making sure your Elasticsearch cluster has enough memory. allow_highlighting = true # settings to be passed to elasticsearch's client (overriding those in the provided elasticsearch_config_file) # all these # this must be the same as for your Elasticsearch cluster elasticsearch_cluster_name = graylog2 # you could also leave this out, but makes it easier to identify the graylog2 client instance elasticsearch_node_name = graylog.test.local # we don't want the graylog2 server to store any data, or be master node #elasticsearch_node_master = false #elasticsearch_node_data = false # use a different port if you run multiple Elasticsearch nodes on one machine #elasticsearch_transport_tcp_port = 9350 # we don't need to run the embedded HTTP server here #elasticsearch_http_enabled = false elasticsearch_discovery_zen_ping_multicast_enabled = false elasticsearch_discovery_zen_ping_unicast_hosts = host1.test.local:9300,host2.test.local:9300,host3.test.local:9300 # Change the following setting if you are running into problems with timeouts during Elasticsearch cluster discovery. # The setting is specified in milliseconds, the default is 5000ms (5 seconds). #elasticsearch_cluster_discovery_timeout = 5000 # the following settings allow to change the bind addresses for the Elasticsearch client in graylog2 # these settings are empty by default, letting Elasticsearch choose automatically, # override them here or in the 'elasticsearch_config_file' if you need to bind to a special address # refer to http://www.elasticsearch.org/guide/en/elasticsearch/reference/0.90/modules-network.html # for special values here #elasticsearch_network_host = #elasticsearch_network_bind_host = #elasticsearch_network_publish_host = # The total amount of time discovery will look for other Elasticsearch nodes in the cluster # before giving up and declaring the current node master. #elasticsearch_discovery_initial_state_timeout = 3s # Analyzer (tokenizer) to use for message and full_message field. The "standard" filter usually is a good idea. # All supported analyzers are: standard, simple, whitespace, stop, keyword, pattern, language, snowball, custom # Elasticsearch documentation: http://www.elasticsearch.org/guide/reference/index-modules/analysis/ # Note that this setting only takes effect on newly created indices. elasticsearch_analyzer = standard # Batch size for the Elasticsearch output. This is the maximum (!) number of messages the Elasticsearch output # module will get at once and write to Elasticsearch in a batch call. If the configured batch size has not been # reached within output_flush_interval seconds, everything that is available will be flushed at once. Remember # that every outputbuffer processor manages its own batch and performs its own batch write calls. # ("outputbuffer_processors" variable) output_batch_size = 5000 # Flush interval (in seconds) for the Elasticsearch output. This is the maximum amount of time between two # batches of messages written to Elasticsearch. It is only effective at all if your minimum number of messages # for this time period is less than output_batch_size * outputbuffer_processors. output_flush_interval = 1 # The number of parallel running processors. # Raise this number if your buffers are filling up. processbuffer_processors = 5 outputbuffer_processors = 3 #outputbuffer_processor_keep_alive_time = 5000 #outputbuffer_processor_threads_core_pool_size = 3 #outputbuffer_processor_threads_max_pool_size = 30 #output_module_timeout = 10000 # UDP receive buffer size for all message inputs (e. g. SyslogUDPInput). #udp_recvbuffer_sizes = 1048576 # Wait strategy describing how buffer processors wait on a cursor sequence. (default: sleeping) # Possible types: # - yielding # Compromise between performance and CPU usage. # - sleeping # Compromise between performance and CPU usage. Latency spikes can occur after quiet periods. # - blocking # High throughput, low latency, higher CPU usage. # - busy_spinning # Avoids syscalls which could introduce latency jitter. Best when threads can be bound to specific CPU cores. processor_wait_strategy = blocking # Size of internal ring buffers. Raise this if raising outputbuffer_processors does not help anymore. # For optimum performance your LogMessage objects in the ring buffer should fit in your CPU L3 cache. # Start server with --statistics flag to see buffer utilization. # Must be a power of 2. (512, 1024, 2048, ...) ring_size = 1024 # Number of threads used exclusively for dispatching internal events. Default is 2. #async_eventbus_processors = 2 # EXPERIMENTAL: Dead Letters # Every failed indexing attempt is logged by default and made visible in the web-interface. You can enable # the experimental dead letters feature to write every message that was not successfully indexed into the # MongoDB "dead_letters" collection to make sure that you never lose a message. The actual writing of dead # letter should work fine already but it is not heavily tested yet and will get more features in future # releases. dead_letters_enabled = false # How many seconds to wait between marking node as DEAD for possible load balancers and starting the actual # shutdown process. Set to 0 if you have no status checking load balancers in front. lb_recognition_period_seconds = 0 # Every message is matched against the configured streams and it can happen that a stream contains rules which # take an unusual amount of time to run, for example if its using regular expressions that perform excessive backtracking. # This will impact the processing of the entire server. To keep such misbehaving stream rules from impacting other # streams, Graylog2 limits the execution time for each stream. # The default values are noted below, the timeout is in milliseconds. # If the stream matching for one stream took longer than the timeout value, and this happened more than "max_faults" times # that stream is disabled and a notification is shown in the web interface. #stream_processing_timeout = 2000 #stream_processing_max_faults = 3 # Length of the interval in seconds in which the alert conditions for all streams should be checked # and alarms are being sent. #alert_check_interval = 60 # Since 0.21 the graylog2 server supports pluggable output modules. This means a single message can be written to multiple # outputs. The next setting defines the timeout for a single output module, including the default output module where all # messages end up. # # Time in milliseconds to wait for all message outputs to finish writing a single message. #output_module_timeout = 10000 # Time in milliseconds after which a detected stale master node is being rechecked on startup. #stale_master_timeout = 2000 # Time in milliseconds which Graylog2 is waiting for all threads to stop on shutdown. #shutdown_timeout = 30000 # MongoDB Configuration mongodb_useauth = true mongodb_user = grayloguser mongodb_password = gray123 mongodb_host = 127.0.0.1 #mongodb_replica_set = localhost:27017,localhost:27018,localhost:27019 mongodb_database = graylog2 mongodb_port = 27017 # Raise this according to the maximum connections your MongoDB server can handle if you encounter MongoDB connection problems. mongodb_max_connections = 100 # Number of threads allowed to be blocked by MongoDB connections multiplier. Default: 5 # If mongodb_max_connections is 100, and mongodb_threads_allowed_to_block_multiplier is 5, then 500 threads can block. More than that and an exception will be thrown. # http://api.mongodb.org/java/current/com/mongodb/MongoOptions.html#threadsAllowedToBlockForConnectionMultiplier mongodb_threads_allowed_to_block_multiplier = 5 # Drools Rule File (Use to rewrite incoming log messages) # See: http://graylog2.org/resources/documentation/general/rewriting #rules_file = /etc/graylog2.drl # Email transport transport_email_enabled = true transport_email_hostname = smtp.test.local transport_email_port = 25 transport_email_use_auth = false transport_email_use_tls = false transport_email_use_ssl = false #transport_email_auth_username = [email protected] #transport_email_auth_password = secret transport_email_subject_prefix = [graylog2] transport_email_from_email = [email protected] # Specify and uncomment this if you want to include links to the stream in your stream alert mails. # This should define the fully qualified base url to your web interface exactly the same way as it is accessed by your users. transport_email_web_interface_url = https://graylog.test.local/ # HTTP proxy for outgoing HTTP calls #http_proxy_uri = http://www-proxy.test.local:8080 # Switch to enable/disable the off-heap message cache. Stores cached messages in the spool directory if set to true. # Stores the messages in an in-memory data structure if set to false. #message_cache_off_heap = true # Directory for the off-heap message cache data. (absolute or relative) #message_cache_spool_dir = spool # The commit interval for the message cache in milliseconds. Only affects message cache implementations that need to commit data. #message_cache_commit_interval = 1000 # When more messages are coming in as we can process, incoming messages will be cached in memory until # they are ready to be processed. Per default this data structure is unbounded, so in situations of # constant high load, it will grow endlessly until all allocatable memory has been consumed and the # process dies. # To prevent this, the next setting allows you to define an upper bound for this memory cache, if the # number of messages in the cache has reached this limit, incoming messages will be dropped until it # has shrunk again. # # The default is 0, which means no upper bound. #input_cache_max_size = 0 # Disable the optimization of Elasticsearch indices after index cycling. This may take some load from Elasticsearch # on heavily used systems with large indices, but it will decrease search performance. The default is to optimize # cycled indices. #disable_index_optimization = true # Optimize the index down to <= index_optimization_max_num_segments. A higher number may take some load from Elasticsearch # on heavily used systems with large indices, but it will decrease search performance. The default is 1. #index_optimization_max_num_segments = 1 # Disable the index range calculation on all open/available indices and only calculate the range for the latest # index. This may speed up index cycling on systems with large indices but it might lead to wrong search results # in regard to the time range of the messages (i. e. messages within a certain range may not be found). The default # is to calculate the time range on all open/available indices. #disable_index_range_calculation = true # Disable the output cache used to buffer messages when the backend storage, e. g. Elasticsearch or any other output # plugin is too slow to handle the amount of messages being indexed. It is recommended to disable the output cache # when using inputs with proper backpressure handling, e. g. Graylog2 radio, AMQP, or Kafka. The default is having # the output cache disabled. disable_output_cache = true # The threshold of the garbage collection runs. If GC runs take longer than this threshold, a system notification # will be generated to warn the administrator about possible problems with the system. Default is 1 second. #gc_warning_threshold = 1s # Connection timeout for a configured LDAP server (e. g. ActiveDirectory) in milliseconds. #ldap_connection_timeout = 2000 # https://github.com/bazhenov/groovy-shell-server #groovy_shell_enable = false #groovy_shell_port = 6789 # Enable collection of Graylog2-related metrics into MongoDB #enable_metrics_collection = false
# Directory where the Elasticsearch binary distribution resides ES_HOME=/usr/share/elasticsearch # Heap Size (defaults to 256m min, 1g max) ES_HEAP_SIZE=4g # Heap new generation #ES_HEAP_NEWSIZE= # max direct memory #ES_DIRECT_SIZE= # Additional Java OPTS #ES_JAVA_OPTS= # Maximum number of open files MAX_OPEN_FILES=65535 # Maximum amount of locked memory #MAX_LOCKED_MEMORY= # Maximum number of VMA (Virtual Memory Areas) a process can own MAX_MAP_COUNT=262144 # Elasticsearch log directory LOG_DIR=/var/log/elasticsearch # Elasticsearch data directory DATA_DIR=/var/lib/elasticsearch # Elasticsearch work directory WORK_DIR=/tmp/elasticsearch # Elasticsearch conf directory CONF_DIR=/etc/elasticsearch # Elasticsearch configuration file (elasticsearch.yml) CONF_FILE=/etc/elasticsearch/elasticsearch.yml # User to run as, change this to a specific elasticsearch user if possible # Also make sure, this user can write into the log directories in case you change them # This setting only works for the init script, but has to be configured separately for systemd startup ES_USER=elasticsearch # Configure restart on package upgrade (true, every other setting will lead to not restarting) #RESTART_ON_UPGRADE=true
