Sasha

Services

Archiver Services

Overview

• The Sasha archiver is responsible for lifecycle management of archived metadata and files for WorkUnits and DFU data.
• It supports both scheduled background archiving and command-driven archive/backup/restore/list flows.
• In bare-metal mode, one combined server runs all archiver roles; in containerized mode, roles are split into dedicated servers.

Main Entry Points

• Server factory declarations: dali/sasha/saarch.hpp
  • createSashaArchiverServer() (bare-metal combined server)
  • createSashaWUArchiverServer() (containerized WU archiver)
  • createSashaDFUWUArchiverServer() (containerized DFU WU archiver)
  • createSashaDFURecoveryArchiverServer() (containerized DFU recovery archiver)
  • createSashaCachedWURemoverServer() (containerized cached WU cleanup)
• Command dispatcher: processArchiverCommand(ISashaCommand *cmd) in dali/sasha/saarch.cpp
• WU listing/query helper used by command actions: WUiterate(ISashaCommand *cmd, const char *mask)

Runtime Topology

1. Bare-metal topology:

   • CSashaCombinedArchiverServer creates and runs:
     • CWorkUnitArchiver
     • CDFUWorkUnitArchiver
     • CDFUrecoveryArchiver
     • CCachedWorkUnitRemover
   • The server loops every minute, checks each archiver schedule, then runs eligible archivers.

2. Containerized topology:

   • CSashaArchiverServer is a base single-role threaded server.
   • Specialized servers each own one archiver instance and one SDS subtree (xpath):
     • CSashaWUArchiverServer -> WorkUnits/
     • CSashaDFUWUArchiverServer -> DFU/WorkUnits/
     • CSashaDFURecoveryArchiverServer -> DFU/RECOVERY/job
     • CSashaCachedWURemoverServer -> CachedWorkUnits/

Core Archiver Framework

• Abstract scheduling/scanning logic is centralized in CBranchArchiver.
• Each concrete branch archiver supplies branch-specific behavior via createBranchItem() and item implementation.
• CBranchArchiver::action() performs:
  1. SDS branch scan (branchxpath)
  2. Per-item qualification (qualifies()) for archive and optional backup (qualifiesBackup())
  3. Sort/prioritization using compare()
  4. Rate/duration control via CSashaSchedule
  5. Archive and backup execution loops

Branch Item Contract

• IBranchItem defines the per-record decision and execution interface:
  • isempty()
  • qualifies()
  • qualifiesBackup()
  • archive()
  • backup()
  • compare(IBranchItem &to)
• Concrete item implementations:
  • CWorkUnitArchiver::cWUBranchItem
  • CDFUWorkUnitArchiver::cDFUWUBranchItem
  • CDFUrecoveryArchiver::cDRBranchItem

WorkUnit Archiving Flow

1. Scheduler path:

   • CWorkUnitArchiver scans WorkUnits/*.
   • Each WU is wrapped by cWUBranchItem.
   • qualifies() gates by protection/time/cutoff and state eligibility.

2. Archive execution:

   • CWorkUnitArchiver::archive() calls doArchiveWorkUnit().
   • doArchiveWorkUnit():
     • Acquires the archive mutex before opening the WU, matching the restore lock order
     • Opens WU via IWorkUnitFactory
     • Optionally version-renames existing LDS files
     • Builds an archive metadata snapshot (when publishing is enabled)
     • Calls IExtendedWUInterface::archiveWorkUnit()
     • Publishes the snapshot only after archiveWorkUnit() succeeds; a publish failure is logged and does not undo the archive

3. Archive publishing (optional):

   • Publisher interface: dali/sasha/wuarchivepublisher.hpp

DFU and Recovery Flows

• DFU WorkUnits:
  • Archiver: CDFUWorkUnitArchiver
  • Archive operation: doArchiveDfuWorkUnit() writes XML to Archive/DFUWorkUnits/...
• DFU Recovery:
  • Archiver: CDFUrecoveryArchiver
  • Archive operation writes XML to Archive/DFUrecovery/... and removes original node (when enabled)
• Cached WU cleanup:
  • CCachedWorkUnitRemover overrides action() to prune older cached entries beyond configured limits.

Command-Driven Flow

• processArchiverCommand() handles interactive/remote command actions:
  • SCA_ARCHIVE, SCA_BACKUP, SCA_RESTORE
    • WU: doArchiveWorkUnit() / doRestoreWorkUnit()
    • DFU: doArchiveDfuWorkUnit() / doRestoreDfuWorkUnit()
  • SCA_LIST, SCA_GET, SCA_WORKUNIT_SERVICES_GET, SCA_LISTDT
    • Delegates to WUiterate()
• Results are appended to command IDs for compatibility with existing consumers.

Synchronization and Error Handling

• The shared CriticalSection (archivingSect) serializes DFU WorkUnit archive/restore operations within the process and is used for WorkUnit archive/restore synchronization.
• Each server thread catches IException, logs via EXCLOG, and requests server stop on unrecoverable errors.
• WorkUnit archive failures mark SDS metadata (@archiveError, @archiveErrorDate) for retry logic.

Configuration Concepts

• Scheduling/limits are read per archiver branch from config (interval, at, limit, cutoff, optional backup).
• WU archiver-specific options include keepResultFiles and retryinterval.
• Archive publishing is provided by a separate publisher abstraction.

Practical File Map

• Archiver interfaces and factories: dali/sasha/saarch.hpp
• Main implementation and orchestration: dali/sasha/saarch.cpp
• Command model: dali/sasha/sacmd.hpp, dali/sasha/sacmd.cpp
• Archive publisher interface: dali/sasha/wuarchivepublisher.hpp
• Archive publisher factory: dali/sasha/wuarchivepublisher.cpp

Debug Plane Housekeeping Server

Overview

• This is a maintenance service that automatically manages files and directories on the debug plane using a configured retention period.
• This is required to control the disc space used by post-mortem and other files on the debug plane.
• This service is designed to execute on a schedule; a single log entry for a summary of any deleted (or failed to be deleted) files or deleted directories will be generated.

Installation and Configuration

• In containerized mode, the Helm chart deploys the service by default; runtime defaults are shown below.

  • For example values.yaml :

    attribute	description	default	required
    disabled	When true, Helm omits this service's deployment.	false	optional
    interval	Minimum interval between debug plane housekeeping expiry checking (in hours, 0 disables).	24	optional
    at	Optional cron schedule for checking expired files and directories.	Unset; interval only	optional
    expiryDays	Default number of days passed that debug plane files or directories are deleted.	14	optional

User Guide

How it works in steps

1. Helm omits the service deployment if disabled is true.
2. The service reads expiryDays and interval from its component configuration.
3. If either value is 0, the service does not start.
4. The debug plane location is read from the component configuration.
5. The service Thread is started.
6. A loop of the following is started until the service is stopped:
   1. A one minute wait is imposed.
   2. If the schedule using at and interval indicates that the housekeeping is required then:
      1. Any exceptions from the following are caught and logged.
         1. The debug plane directory list is read.
         2. The list of files and directories in the debug plane are processed:
            1. If the file or directory's modified date is over expiryDays old then:
               1. The file or directory is deleted.
               2. A counter for each file or directory is maintained.
         3. A log entry is made giving the total number of files and directories deleted.

Inputs and Outputs

• The component configuration as per Installation and Configuration is the input for this service.
• The outputs are any logs generated for errors and deletion statistics.

Error Handling

• Known exceptions are handled to prevent service interruption.

Limitations and Considerations

• The service ensures it only runs in containerized environments.

Future Possible Enhancements

• A bare metal implementation.
• Logging freed up disc space statistics.