package RAMCloud.ProtoBuf

Get desktop application:
View/edit binary Protocol Buffers messages

This message is used to store overall state about the CoordinatorClusterClock class in external storage.

• required uint64 next_safe_time = 1

  / The Cluster Time that is guaranteed to be greater than or equal to any / value the CoordinatorClusterClock will externalize to anyone, so that in / the event of a coordinator crash, the new coordinator can ensure the / monotonic property of Cluster Time. This time is periodically moved / forward so so that the CoordinatorClusterClock can continue to progress.

This protocol buffer message is used to store persistent data needed by CoordinatorUpdateManager. These protocol buffers are used for storing and retrieving data from ExternalStorage.

• required uint64 lastFinished = 1

  All operations with sequence numbers less than or equal to this value are known to have completed. After a coordinator crash, this tells the new coordinator which operations it must check to ensure that they have completed successfully.

• required uint64 firstAvailable = 2

  Sequence numbers less than this value may have been assigned already, so a new coordinator should not use values smaller than this. Used to ensure that sequence numbers don't get reused after a coordinator crash.

/ The network format for RAMCloud::EnumerationIterator. Any changes / made here must also be made there as well.

• repeated EnumerationIterator.Frame frames = 1

  / See RAMCloud::EnumerationIterator::frames.

/ See RAMCloud::EnumerationIterator::Frame.

Used in: EnumerationIterator

• required uint64 tablet_start_hash = 1

  / See RAMCloud::EnumerationIterator::Frame::tabletStartHash.

• required uint64 tablet_end_hash = 2

  / See RAMCloud::EnumerationIterator::Frame::tabletEndHash.

• required uint64 num_buckets = 3

  / See RAMCloud::EnumerationIterator::Frame::numBuckets.

• required uint64 bucket_index = 4

  / See RAMCloud::EnumerationIterator::Frame::bucketIndex.

• required uint64 bucket_next_hash = 5

  / See RAMCloud::EnumerationIterator::Frame::bucketNextHash.

/ Serialization of a Histogram object. Please see Histogram.h's member / variable documentation for details on the fields below.

Used in: LogMetrics.CleanerMetrics.OnDiskMetrics, LogMetrics.SegmentMetrics

• required fixed64 num_buckets = 1

• required fixed64 bucket_width = 2

• repeated Histogram.Bucket bucket = 3

• required fixed64 sample_sum_high = 4

• required fixed64 sample_sum_low = 5

• required fixed64 outliers = 6

• required fixed64 max = 7

• required fixed64 min = 8

Each pair stores a unique bucket index and the count. Unclear if doing this an omitting empty buckets will be a net win over also including the index value.

Used in: Histogram

• required fixed64 index = 1

• required fixed64 count = 2

Used in: RecoveryPartition

• required uint64 table_id = 1

  /The id of the containing table

• required uint32 index_id = 2

  /The id of the index

• required uint64 backing_table_id = 3

  /The id of the table that will hold objects for this indexlet.

• optional bytes first_key = 4

  /The first (smallest) key that is part of this indexlet.

• optional bytes first_not_owned_key = 5

  /The smallest key greater than all the keys managed by this indexlet.

• optional fixed64 server_id = 6

  / The id of the master owning this indexlet.

• optional string service_locator = 7

  / The service locator for the master owning this indexlet.

• optional fixed64 user_data = 8

  / User data

/ Various metrics regarding log space usage and cleaner performance. These / fields are accumulated from a number of classes, including Log, LogCleaner, / LogSegment, SegletAllocator, and SegmentManager. / / Clients can query a server's log for these metrics in order to observe what / sort of load it's under and how it's performing.

• required double ticks_per_second = 1

  / Frequency of the server's timestamp counter. Needed to properly convert / the advertised ticks into seconds across machines of different speed. / The machines should probably just return nanosecond values instead, but / we can fix that another time.

• required fixed64 total_sync_calls = 2

  / The following fields are metrics maintained by the (Abstract)Log / classes. They're documented in Log::Metrics and AbstractLog::Metrics.

• required fixed64 total_sync_ticks = 3

• required fixed64 total_append_calls = 4

• required fixed64 total_append_ticks = 5

• required fixed64 total_no_space_ticks = 6

• required fixed64 total_bytes_appended = 7

• required fixed64 total_metadata_bytes_appended = 8

• required LogMetrics.CleanerMetrics cleaner_metrics = 9

• required LogMetrics.SegletMetrics seglet_metrics = 10

• required LogMetrics.SegmentMetrics segment_metrics = 11

/ Log metrics related to cleaning. Filled in by the LogCleaner class.

Used in: LogMetrics

• required fixed32 poll_usec = 1

  / The following are compile-time constants. See LogCleaner.h for / details on their meaning.

• required fixed32 max_cleanable_memory_utilization = 2

• required fixed32 live_segments_per_disk_pass = 3

• required fixed32 survivor_segments_to_reserve = 4

• required fixed32 min_memory_utilization = 5

• required fixed32 min_disk_utilization = 6

• required fixed64 do_work_ticks = 7

• required fixed64 do_work_sleep_ticks = 8

• required CleanerMetrics.InMemoryMetrics in_memory_metrics = 9

• required CleanerMetrics.OnDiskMetrics on_disk_metrics = 10

• required CleanerMetrics.ThreadMetrics thread_metrics = 11

/ Serialized form of LogCleanerMetrics::InMemory. See the C++ class / documentation for details.

Used in: CleanerMetrics

• required fixed64 total_relocation_callbacks = 1

• required fixed64 total_relocation_appends = 2

• required fixed64 total_bytes_freed = 3

• required fixed64 total_bytes_in_compacted_segments = 4

• required fixed64 total_bytes_appended_to_survivors = 5

• required fixed64 total_segments_compacted = 6

• required fixed64 total_empty_segments_compacted = 7

• repeated fixed64 total_entries_scanned = 8

  / The index of each count corresponds to the LogEntryType enum.

• repeated fixed64 total_live_entries_scanned = 9

• repeated fixed64 total_scanned_entry_lengths = 10

• repeated fixed64 total_live_scanned_entry_lengths = 11

• required fixed64 total_ticks = 12

• required fixed64 get_segment_to_compact_ticks = 13

• required fixed64 wait_for_free_survivor_ticks = 14

• required fixed64 relocation_callback_ticks = 15

• required fixed64 relocation_append_ticks = 16

• required fixed64 compaction_complete_ticks = 17

/ Serialized form of LogCleanerMetrics::OnDisk. See the C++ class / documentation for details.

Used in: CleanerMetrics

• required fixed64 total_bytes_appended_to_survivors = 1

• required fixed64 total_memory_bytes_freed = 2

• required fixed64 total_disk_bytes_freed = 3

• required fixed64 total_memory_bytes_in_cleaned_segments = 4

• required fixed64 total_disk_bytes_in_cleaned_segments = 5

• required fixed64 total_relocation_callbacks = 6

• required fixed64 total_relocation_appends = 7

• required fixed64 total_segments_cleaned = 8

• required fixed64 total_empty_segments_cleaned = 9

• required fixed64 total_survivors_created = 10

• required fixed64 total_runs = 11

• required fixed64 total_low_disk_space_runs = 12

• required fixed64 memory_utilization_at_start_sum = 13

• repeated fixed64 total_entries_scanned = 14

  / The index of each count corresponds to the LogEntryType enum.

• repeated fixed64 total_live_entries_scanned = 15

• repeated fixed64 total_scanned_entry_lengths = 16

• repeated fixed64 total_live_scanned_entry_lengths = 17

• required fixed64 total_ticks = 18

• required fixed64 get_segments_to_clean_ticks = 19

• required fixed64 cost_benefit_sort_ticks = 20

• required fixed64 get_sorted_entries_ticks = 21

• required fixed64 timestamp_sort_ticks = 22

• required fixed64 relocate_live_entries_ticks = 23

• required fixed64 wait_for_free_survivors_ticks = 24

• required fixed64 cleaning_complete_ticks = 25

• required fixed64 relocation_callback_ticks = 26

• required fixed64 relocation_append_ticks = 27

• required fixed64 close_survivor_ticks = 28

• required fixed64 survivor_sync_ticks = 29

• required Histogram cleaned_segment_memory_histogram = 30

• required Histogram cleaned_segment_disk_histogram = 31

• required Histogram all_segments_disk_histogram = 32

/ Serialized form of LogCleanerMetrics::Threads. See the C++ class / documentation for details.

Used in: CleanerMetrics

• repeated fixed64 active_ticks = 1

/ Log metrics related to allocation of seglets to in-memory segments. / Filled in by the SegletAllocator class.

Used in: LogMetrics

• required fixed64 total_seglets = 1

  / Total number of seglets in the system, including reserves.

• required fixed64 total_usable_seglets = 2

  / Total number of seglets usable for storing data via log appends.

• required fixed64 emergency_head_pool_reserve = 3

  / Number of seglets reserved for emergency head segments. These allow / the cleaner to issue a new digest and free space even when under / heavy memory pressure.

• required fixed64 emergency_head_pool_count = 4

  / Number of seglets left in the emergency head reserve. If less than / the reserve amount, one or more emergency heads must currently be / allocated.

• required fixed64 cleaner_pool_reserve = 5

  / Number of seglets reserved for the log cleaner.

• required fixed64 cleaner_pool_count = 6

  / Number of seglets left in the cleaner reserve. If less than the / reserve amount, some are currently allocated for cleaning.

• required fixed64 default_pool_count = 7

  / Number of seglets available for storing data in new head segments.

/ Log metrics related to allocated segments, both in memory and on disk. / Filled in by the SegmentManager class.

Used in: LogMetrics

• required Histogram segments_on_disk_histogram = 1

  / Histogram tracking the number of segments replicated on backups. / Note that this value does not take R into account, so it is not / the number of total replicas.

• required fixed64 current_segments_on_disk = 2

  / Current count of the number of segments allocated on backups.

• repeated fixed64 total_entry_counts = 3

  / The index of each count corresponds to the LogEntryType enum.

• repeated fixed64 total_entry_lengths = 4

* Log metadata for a master which is stored on the coordinator for use during its recovery. This is expected to be small and updated extremely rarely. Currently, masters use this to invalidate replicas which may have become inconsistent due to backup crashes. Even in an extremely large cluster this is expected to happen a few times per hour. When a backup crashes that a master was replicating an open segment to the ReplicaManager increments an "epoch" number for that segment. Replicas for that segment are tagged with the new epoch number (until enough of them are created to meet the replication factor). To prevent old replicas that were created with the old epoch number from being used it stores that epoch number here. Only the epoch for the most recent segment to experience a failure while open is stored. That is because, by construction, no earlier segment should be considered open/the head anyway during recovery (they must be closed). Therefore, any replica found on backups with a segment id less than what is included in this struct is invalid. Likewise, any replica found with a matching segment id but a lesser epoch is also invalid. Invalid replicas must not be used by recovery in any way: that includes for replay AND log digest.

Used in: ServerListEntry

• optional uint64 min_open_segment_id = 1

  * Minimum segment id which a replica must have to be a valid replica of the log *IF* the replica is open (subject to the constraint of the #min_open_segment_epoch below). Closed replicas are always valid. Note: making this optional with a default value avoids the need to explicitly initialize in CoordinatorServerList.cc, so it doesn't have to understand what's in these protobufs.

• optional uint64 min_open_segment_epoch = 2

  * Minimum epoch a replica must have *IF* the replica is open *AND* the segment id is exactly the same as above. See note above about optional/default.

/ A list containing one or more performance counters from a RAMCloud / server.

• repeated MetricList.Entry metric = 1

  / The full list of metrics.

/ An individual metric.

Used in: MetricList

• required string name = 1

  / The name for this metric, such as "master.replicas".

• required uint64 value = 2

  / Current value for the metric.

Message used by coordinator and recoveryMaster during recovery. At the beginning of a recovery, coordinator send a recoveryPartition to each recoveryMaster. When recoveryMaster finishes its job, it fills in some fields of recoveryPartition, and sends it back to coordinator. Same message goes both directions during a recovery.

• repeated Tablets.Tablet tablet = 1

  / The tablets.

• repeated Indexlet indexlet = 2

  / The indexlets.

/ Configuration details for a server. This is a serialized form of all fields / in the ServerConfig class.

• required string coordinator_locator = 1

  / A locator the server can use to contact the cluster coordinator.

• required string local_locator = 2

  / The locator the server should listen for incoming messages on.

• required string cluster_name = 3

  / Tag given to replicas stored on this backup. See the ServerConfig class / documentation for more details.

• required string services = 4

  / Which services this server is running.

• required bool detect_failures = 5

  / Whether the failure detection thread should be started.

• required bool pin_memory = 6

  / All memory in the process is pinned.

• required fixed32 segment_size = 7

  / Size of each segment in bytes.

• required fixed32 seglet_size = 8

  / Size of each seglet in bytes.

• required fixed32 max_object_data_size = 9

  / Largest allowable RAMCloud object, in bytes.

• required fixed32 max_object_key_size = 10

  / Largest allowable key for a RAMCloud object, in bytes.

• required fixed32 max_cores = 11

  / Max number of cores to use at once for dispatch and worker threads.

• optional ServerConfig.Master master = 12

  / The server's MasterService configuration, if it is running one.

• optional ServerConfig.Backup backup = 13

  / The server's BackupService configuration, if it is running one.

Used in: ServerConfig

• required bool gc = 1

  / Whether the backup garbage collector is enabled or not.

• required bool in_memory = 2

  / Whether the BackupService should store replicas in RAM or on disk.

• required fixed32 num_segment_frames = 3

  / Number of replica-sized storage chunks to allocate on the backup's / backing store.

• required fixed32 max_non_volatile_buffers = 4

  / Maximum number of segments that the backup will buffer in memory / before rejecting new segments until the disk catches up.

• required fixed32 max_recovery_replicas = 20

  / Maximum number of replicas that any given master recovery will / keep in memory at any given time.

• optional string file = 5

  / Path to a file to use for the backing store if inMemory is false.

• required int32 strategy = 6

  / BackupStrategy to use for balancing replicas across backups.

• required fixed32 mock_speed = 7

  / If non-0, the fake backup disk speed this server is advertising.

• required fixed64 write_rate_limit = 8

  / If non-0, limit writes to backup to this many megabytes per second.

/ Configuration details specific to the MasterService on a server.

Used in: ServerConfig

• required fixed64 log_bytes = 1

  / Total number bytes to use for the in-memory Log.

• required fixed64 hash_table_bytes = 2

  / Total number of bytes to use for the HashTable.

• required bool disable_log_cleaner = 3

  / If true, disable the log cleaner entirely.

• required bool disable_in_memory_cleaning = 4

  / If true, disable in-memory cleaning and always do disk cleaning.

• required double backup_disk_expansion_factor = 5

  / Determines how much space may be used on backup disks, as a multiple / of the in-memory log size.

• required fixed32 cleaner_write_cost_threshold = 6

  / If in-memory cleaning is enabled, this specifies the balance between / in-memory and disk cleaning.

• required string cleaner_balancer = 7

  / LogCleaner::Balancer type used to schedule cleaning and compaction.

• required fixed32 cleaner_thread_count = 8

  / Maximum number of threads the cleaner will use.

• required fixed32 num_replicas = 9

  / Number of replicas to keep per segment stored on backups.

• required bool use_hugepages = 10

  / Specifies whether to use hugepage memory to allocate / LargeBlockOfMemory.

• required bool use_mincopysets = 11

  / Specifies whether to use MinCopysets or random replication.

• required bool use_local_backup = 12

  / If true, allow replication to local backup.

• required bool use_plusonebackup = 13

  / Specifies whether to use masterServerId plus one with wraparound / or random replication for backupServerId.

/ A list of servers, which describes the services they provide and a / means of contacting them using RAMCloud transports. / / This message is used both when pulling complete lists from the / coordinator (GetServerListRpc) and when pushing delta updates / to other servers by the coordinator (UpdateServerListRpc).

• repeated ServerList.Entry server = 1

  / List of servers. If type is UPDATE then this list should only have / a single entry.

• required fixed64 version_number = 2

  / Generation number of the Coordinator's list that corresponds to / the information in this message. Used to ignore duplicate or / oout-of-order information. See CoordinatorServerList::version and / ServerList::version for more details.

• required ServerList.Type type = 3

Used in: ServerList

• required fixed32 services = 1

  / Which services this server is running.

• required fixed64 server_id = 2

  / A unique id for the service.

• required string service_locator = 3

  / A locator for the service.

• required fixed32 expected_read_mbytes_per_sec = 5

  / Disk bandwidth of the backup server in MB/s, if this entry represents / a server which is running a backup service.

• required fixed32 status = 6

  / The current status of this server in terms of cluster membership. / See above for ServerStatus in ServerList.h for legal enum values.

• required fixed64 replication_id = 7

  / The replicationId of the server.

Used in: ServerList

• FULL_LIST = 0

  / Specifies that this is a full server list

• UPDATE = 1

  / Specifies that this is a delta update

/ Describes the contents of a single entry in a CoordinatorServerList. / Used to save information on external storage in order to recover / the CoordinatorServerList after a coordinator crash.

• required fixed32 services = 1

  / Which services this server is running.

• required fixed64 server_id = 2

  / Unique identifier for the service (contents of a ServerId object).

• required string service_locator = 3

  / Locator for the service.

• required fixed32 expected_read_mbytes_per_sec = 4

  / Disk bandwidth of the backup server in MB/s, if this entry represents / a server which is running a backup service.

• required fixed32 status = 5

  / The current status of this server in terms of cluster membership. / See ServerStatus enum in AbstractServerList.h for legal values.

• required fixed64 replication_id = 6

  / Identifies the replication group containing this server.

• required MasterRecoveryInfo master_recovery_info = 7

  / Information needed for master recovery, which is managed outside / CoordinatorServerList; we store it and recover it without knowing / what it is or exactly how it's used.

• repeated ServerListEntry.Update update = 8

  / Updates that may not have been received by all of the servers / in the cluster yet. There will always be at least one element / in this list (the most recent update).

/ Describes one update to this entry

Used in: ServerListEntry

• required fixed32 status = 1

  / New status of the server, as of this update. See ServerStatus enum / for legal values.

• required uint64 version = 2

  / Server list version corresponding to this update.

• required uint64 sequence_number = 3

  / CoordinatorUpdateManager's sequence number for this update.

/ A list of statistical information about a single master server. / / This message is used when a master server is asked for its statistical / information via GetServerStatisticsRpc.

• repeated ServerStatistics.TabletEntry tabletentry = 1

  / List of TabletEntries.

• required SpinLockStatistics spin_lock_stats = 2

  / Stats on all SpinLock instances, to monitor contention.

Each tablet on the master has a correspsonding TabletEntry in the ServerStatistics where its access statistics are being stored.

Used in: ServerStatistics

• required uint64 table_id = 1

  / The id of the containing table.

• required uint64 start_key_hash = 2

  / The smallest hash value for a key that is in this tablet.

• required uint64 end_key_hash = 3

  / The largest hash value for a key that is in this tablet.

• optional uint64 number_read_and_writes = 4

  / Read and write access statistics for a single tablet.

/ A list of SpinLocks instantiated in a server, including the lock's / name (if one was given), and statistics used to monitor contention / for each lock.

Used in: ServerStatistics

• repeated SpinLockStatistics.Lock locks = 1

/ See SpinLock.h.

Used in: SpinLockStatistics

• required string name = 1

  / String name of the SpinLock.

• required fixed64 acquisitions = 2

  / Number of times the lock has been acquired.

• required fixed64 contended_acquisitions = 3

  / Number of times the lock has been acquired, but only after having / to spin wait first.

• required fixed64 contended_nsec = 4

  / Total number of nanoseconds spent waiting to acquire the lock when / it was already held.

For each table, the coordinator stores one of these messages in external storage. The message contains two things: (a) the current state of the table, and (b) information about operations in progress, which will used to complete those operations if the current coordinator crashes.

• required string name = 1

  / The table's name, as specified in the createTable RPC that / created the table.

• required uint64 id = 2

  / Identifier for the table (used in RPCs such as read and write).

• repeated Table.Tablet tablet = 3

  / The tablets.

• required uint64 sequence_number = 4

  / Sequence number for the most recent update. 0 means the most / recent update does not have any recovery-related issues.

• optional bool created = 5

  / If this value is present, it is true, indicating that the most / recent update was to create the table.

• optional bool deleted = 6

  / If this value is present, it is true, indicating that the most / recent update was to delete the table.

• optional Table.Split split = 7

• optional Table.Reassign reassign = 8

• repeated Table.Index index = 9

• optional Table.ReassignIndexlet reassign_indexlet = 10

/ Each message of this type holds information about an index corresponding / to this table.

Used in: Table

• required uint32 index_id = 1

  / ID for the particular secondary index.

• repeated Index.Indexlet indexlet = 2

/ Each message of this types holds information about an indexlet / that is part of this index.

Used in: Index

• required string first_key = 1

  / Key blob marking the start of the indexed key range for this indexlet.

• required string first_not_owned_key = 2

  / Blob of the smallest key in the given index that is after firstKey / in the index order but not part of this indexlet.

• required uint64 backing_table_id = 3

  / ID of the backing table that will hold objects for this indexlet. / This table will have only one tablet that is entirely owned by one / server. It will never be partitioned into tablets across different / servers.

• required fixed64 server_id = 4

  / ID of the master owning the table with ID backing_table_id.

/ A message of this type is present if the most recent update was / to reassign a tablet.

Used in: Table

• required fixed64 server_id = 1

  / ID of the server to which the tablet was reassigned.

• required uint64 start_key_hash = 2

  / Key hash range for the reassigned tablet.

• required uint64 end_key_hash = 3

/ A message of this type is present if the most recent update was / to reassign an indexlet.

Used in: Table

• required fixed64 server_id = 1

  / ID of the server to which the indexlet was reassigned.

• required uint32 index_id = 2

  / ID of the secondary index to which this indexlet belongs.

• required bytes first_key = 3

  / Key range for this indexlet.

• required bytes first_not_owned_key = 4

• required uint64 backing_table_id = 5

  / ID of the backing table that will hold objects for this indexlet. / This table will have only one tablet that is entirely owned by one / server. It will never be partitioned into tablets across different / servers.

/ A message of this type is present if the most recent update was / to split a tablet in the table.

Used in: Table

• required fixed64 server_id = 1

  / The ID of the master owning the tablet that was split.

• required uint64 split_key_hash = 2

  / Key hash value of the split point.

/ Each message of this type holds information about one tablet in the / table.

Used in: Table

• required uint64 start_key_hash = 1

  / The smallest hash value for a key that is in this tablet.

• required uint64 end_key_hash = 2

  / The largest hash value for a key that is in this tablet.

• required Tablet.State state = 3

  / The state of the tablet, see above.

• required fixed64 server_id = 4

  / The ID of the master owning this tablet.

• required uint64 ctime_log_head_id = 5

  / Id of the head segment in the log of server_id at the time the / tablet was assigned to the server. Any earlier segments in the / master's log cannot contain data belonging to this tablet.

• required uint32 ctime_log_head_offset = 6

  / Offset in the head segment of server_id's log at the time it was / assigned this tablet. Any objects appearing earlier in that segment / cannot contain data belonging to this tablet.

Used in: Tablet

• NORMAL = 0

  / The tablet is available.

• RECOVERING = 1

  / The tablet is being recovered, so it's not available.

• BOGUS = 2

  / Used only for testing.

Message used for encoding tablet and index information for a particular table id. Used by the object finder to fetch information from coordinator.

• repeated TableConfig.Tablet tablet = 1

  / The tablets.

• repeated TableConfig.Index index = 2

  / The indexes.

Used in: TableConfig

• required uint32 index_id = 1

  / The id of the index.

• required uint32 index_type = 2

  / The type of the index.

• repeated Index.Indexlet indexlet = 3

  / The indexlets for each index.

Used in: Index

• required bytes start_key = 1

  /The first (smallest) key that is part of this indexlet.

• required bytes end_key = 2

  /The smallest key greater than all the keys managed by this indexlet.

• required fixed64 server_id = 3

  / The ID of the master owning this indexlet.

• required string service_locator = 4

  / The service locator for the master owning this indexlet.

Used in: TableConfig

• required uint64 table_id = 1

  / The id of the containing table.

• required uint64 start_key_hash = 2

  / The smallest hash value for a key that is in this tablet.

• required uint64 end_key_hash = 3

  / The largest hash value for a key that is in this tablet.

• required Tablet.State state = 4

  / The state of the tablet, see above.

• optional fixed64 server_id = 5

  / The ID of the master owning this tablet.

• optional string service_locator = 6

  / The service locator for the master owning this tablet.

• optional fixed64 user_data = 7

  / An opaque field which happens to be large enough for a pointer.

• required uint64 ctime_log_head_id = 8

  / Id of the the log head belonging to the master that owns this / tablet when it was assigned to the server. Any earlier segments / cannot contain data belonging to this tablet.

• required uint32 ctime_log_head_offset = 9

  / Offset in the log head segment belonging to the master that owns this / tablet when it was assigned to the server. Any objects appearing / earlier in that segment cannot contain data belonging to this tablet.

Used in: Tablet

• NORMAL = 0

  / The tablet is available.

• RECOVERING = 1

  / The tablet is being recovered, so it's not available.

This message is used to store overall state about the TableManager class on external storage.

• required uint64 next_table_id = 1

  / The smallest table id that is guaranteed never to have been used.

A set of tablets (contiguous ranges of tables).

• repeated Tablets.Tablet tablet = 1

  / The tablets.

Used in: RecoveryPartition, Tablets

• required uint64 table_id = 1

  / The id of the containing table.

• required uint64 start_key_hash = 2

  / The smallest hash value for a key that is in this tablet.

• required uint64 end_key_hash = 3

  / The largest hash value for a key that is in this tablet.

• required Tablet.State state = 4

  / The state of the tablet, see above.

• optional fixed64 server_id = 5

  / The ID of the master owning this tablet.

• optional string service_locator = 6

  / The service locator for the master owning this tablet.

• optional fixed64 user_data = 7

  / An opaque field which happens to be large enough for a pointer.

• required uint64 ctime_log_head_id = 8

  / Id of the the log head belonging to the master that owns this / tablet when it was assigned to the server. Any earlier segments / cannot contain data belonging to this tablet.

• required uint32 ctime_log_head_offset = 9

  / Offset in the log head segment belonging to the master that owns this / tablet when it was assigned to the server. Any objects appearing / earlier in that segment cannot contain data belonging to this tablet.

Used in: Tablet

• NORMAL = 0

  / The tablet is available.

• RECOVERING = 1

  / The tablet is being recovered, so it's not available.

• required string foo = 1