dCache.Org eagle
black_bg
home | documentation | downloads | feedback | search | imprint
black_bg


The 1.8.0-release

dCache 1.8.0 production.
Please make yourself familiar on how to configure Space Reservation, part of the SRM chapter of the the dCache book.


The change log summary

dCache Releases of 1.8.0 (dCache Core Server, SRM + SRM 2.2)

OS, Platform independent

Please find the currently recommended system in green.
Download Rev. Date md5 hash Change Log/Components to upgrade
dCache server (patch 15p7) 23.06.2008 2e025f9f792980e9a89baa5d038880b1 ftp doors, pools, gPlazma
dCache server (patch 15p7) tgz 23.06.2008 d544b3e11f0f88b8868ec0ae715d7c8f
dCache server (patch 15p6) rpm 17.06.2008 752984135c0c8ec757557886fbf6d91e SRM, pools, ftp doors
dCache server (patch 15p6) tgz 17.06.2008 a7567197bc8846ba70aeb72d23b8af05
dCache server (patch 15p5) 3.06.2008 dac4751245e1b4a7f6ad24e6774c317b srm, pools, http, ftp door
dCache server (patch 15p5) tgz 3.06.2008 e432fdbc2e8df56caec0bc5a91ebc76e

dCache Releases of 1.8.0 Client SRM

OS Components Download Rev. Date md5 hash
srmclient (rpm) SRMv2.2 client dCache SRM client (patch level 15p7) 23.06.2008 b584b5d5e4fe31a9f5744d88273ca983
srmclient (rpm) SRMv2.2 client dCache SRM client (patch level 14) 28.02.2008 6c914bcb5e27353c60fe5f0d2fe2663a

dCache Releases of 1.8.0 Client DCAP

OS Components Download Rev. Date md5 hash
dcapclient (rpm) DCAP client x86_64 dCache DCAP client (patch level 4) 22.11.2007 a4b568256bbf0f71916adb18d16a7dc7
dcapclient (rpm) DCAP client i586 dCache DCAP client (patch level 4) 22.11.2007 4107d5eed0d4c9065284cbedb8be8701

The change log summary


ChangeLog

15 => 15p3

SRM, SrmSpaceManager and PinManager (utility Domain) only (changelog)

14 => 15

Synchronized update of SpaceManager and PoolManager required
changelog

13 => 14

This release requires to upgrade all components at the same time (unless you run 1.8.0-13).
  1. various fixes in the new startup scripts
  2. cleaning up install.sh: added logging, moved from sh to bash
  3. fixed socked adapter
  4. new release of Chimera
  5. fixed FTP protocol violation in ftp door, which causes remote gsiftp transfers to fail.
  6. added option to steer number of concurrent removes on SRM remove.
  7. ls in stace manager includes more options
  8. More changes.

12 => 13

This release requires to upgrade all components at the same time.
  1. All changes of the ccrc08 branch have been applied to the release as well.
  2. More changes.

10 => 12

This release requires a space manager database schema migration
  1. An improved SrmSpaceManager has been introduced, which requires a schema migration.
  2. Space manager schema migration script added.
  3. Improved addCacheLocation in PnfsManager.
  4. Improved ftp door
  5. Fixed file descriptor leak in ActiveAdapter
  6. Inter-cell communication uses 'keep alive' sockets, which should fix domain reconnect problems.
  7. make srm copy version 1 to version 2 in push mode specify TargetStorageType=PERMANENT, make srmCopy and srmPrepareToPut reject non-PERMANENT storage types, when copies are into dCache
  8. Added pnfs2chimera migration scripts.
  9. Fixed sticky bit leaking in pin manager.
  10. Added user's DN to remoteGsiTransfers billing record
  11. With this release, the (new) improved PinManager is becoming the default.
  12. Fixed the 'pnfs register' command such that it does not register files that are not marked either precious or cached yet.

9 => 10

  1. Fixed several space management related bugs in pools (although their likely are more bugs in this area - don't assume all problems are fixed!).
  2. Restructured pool inventory generation. If pool startup fails during the inventory phase, a pool is now disabled in such a way that it cannot be enabled with 'pool enable' (as that would lead to further errors).
  3. Removed the -modify flag to pools (it was not fully implemented and was dangerous to use).
  4. Improved exception handling in pools.
  5. Added periodic consistency checks for space accounting inconsistencies. If inconsistencies are detected, a pool is automatically marked as read-only to protect against further degradation of the pool.
  6. Fixed bugs in handling of URI based HSM scripts when used with Berkeley DB based pools.
  7. Fixed race condition and error handling in p2p client.
  8. Fixed bug in pool repair code used during pool startup, which would previously propagate a bad file size to PNFS.
  9. Added command for listing the named sticky bits in the repository file in the pool, and for creating/modifying sticky bits and their lifetimes.
  10. added printouts to trace space allocations
  11. added error report in case of entry.getSize() not equal to entry.getDataFile.length() on cache repository entry removeal
  12. added start/stop/restart to chimera-based nfs interface
  13. added namespace migration utility
  14. set gap command in the pool supports unit (like 5G )
  15. fixed leading zero in adler32 checksum: 675af7 == 00675af7
  16. gsidcap sends VOinfo to space manager if enabled
  17. mover queues can reuse working thread ( thread pool instead of new Thread)
  18. added execution bit for shell scripts in dcache-server-X.tgz
  19. no hidden on-the-fly checksum computing when switched off on the target pool of a P2P-transfer
  20. newer xrootd clients are working against dCache (the mover can now handle asyncronous requests)

8 => 9

This release requires synchronous update of SRM and PoolManager. Other components may run mixed.
  1. Fixed dead lock in the pool
  2. Fixed pool space miss calculation
  3. added jar files used by replica manager
  4. permission hadler replaced with a new implementation
  5. intenal improvements

6 => 8

  1. Added some safety checks to the GFtp/2 mover to protect against potential problems in the data channel implementations. It does so by adjusting the space reservation after the transfer.
  2. Older releases freed all reserved space when something goes wrong with the space manager. This is not correct when a file is left on the pool, and can lead to pool overbooking.
  3. All pools participating in replication MUST be declared in PoolManager.conf in ResilientPools pool group or other group presented to Resilient Manager.
  4. DB schema has changed and must be re-initialized by provided template:
    $ dropdb replicas
    $ createdb replicas
    $ psql -U enstore replicas < /opt/d-cache/etc/psql_install_replicas.sql
    
  5. Pool code changed back so PnfsAddCacheLocationMessage is sent out when file added to the pool. This resolves the issue when files been in receiving.transfer state not replicated.
  6. Files not found in pnfs now correctly excluded from replications.
  7. One more pnfs error report signature was added to exclude files removed from pnfs.
  8. Flags "all pools are resilient" ('*') are removed from templates replica.batch and dCacheSetup.template. This is not supported for sometime as it proved to prone to configuration errors, we removed it now from default configuration.
  9. replica manager : auto release of excluded files after 12 hour timeout. Configurable via option ExcludeExpirationTO.
  10. replica manager : All ongoing replications when pool goes down are stopped.
  11. replica manager : connection to DB is split using DB connection pooling
  12. replica manager : resilient manager monitoring now reports reason for the file exclusion.

5 => 6

  1. space token for a file is retrieved independent on the state of the file
  2. possible bottlneck having to do with sequence generation
  3. proper error message on bad pnfs config
  4. clenaup of "Unexpected reply" in pnfsDomain.log
  5. srmLs showes ONLINE/NEARLINE
  6. pnfsManager do not ties to read level2 on directories
  7. srmLs showes space tocken of a file

4 => 5

  1. fixed broken in -4 directroy creation

3 => 4

  1. fixed unpinning of ONLINE files after transfer
  2. fixed 'binary' compatibility with 1.8.0-2
  3. fixed billingDB error in flush/restore logging
  4. fixed retry blacklist in PoolManager
  5. added restore suspend if hsmcp exit code is 71
  6. added chimera-based nfs server ( Evaluation only )

2 => 3

  1. improved SrmSpaceManager performance
  2. added logs for pool up messages
  3. added pool repository tests
  4. fixed flushing of files which left prior migration from 1.7
  5. fixed active flush counter
  6. added history capabilities into admin interface

1 => 2

  1. detect that the ProtocolInfo is GridProtocolInfo in SelectPool msg, take VO Group and role from it, take Access Latency and RetentionPolicy from storage info for implicit space reservation for non-srrm transfers
  2. added LinkGroup into match page

0 => 1

  1. Get rid of orphaned files on pools after a gridftp/dcap fails.
  2. Correcting statistics in PnfsManager. Now all messages should be counted.
  3. Removed the batchfile-parameter "calculate-transfer-crc'. Checksumming for DCap now entirely steered by the checksum module policy.
  4. /usr/bin/java now default path in dCacheSetup. (We need 1.5 for dCache 1.8 anyway)
  5. All the technical details.

dCache beta Releases of 1.8.0

OS Components Download Rev. Date md5 hash
Platform independent (rpm) dCache Core Server,  SRM (SRM 2.2) dCache server (patch level 26) 25.10.2007 56d9b302db9b30a59b2bb44c321e8321
Platform independent (tgz) dCache Core Server,  SRM (SRM 2.2) dCache server (patch level 26) 25.10.2007 87e1cba29f70566d62c0c2f4c8430819
Platform independent (rpm) dCache Core Server,  SRM (SRM 2.2) dCache server (patch level 23) 22.10.2007 cf461eabf7f4b529026ee20255e2a7f8
Platform independent (tgz) dCache Core Server,  SRM (SRM 2.2) dCache server (patch level 23) 22.10.2007 99b4d00e0c570a73e30db030908d3090
Platform independent (rpm) SRMv2.2 client dCache server (patch level 7)    


ChangeLog

beta25 => beta26

  1. You need to change your configuration in order to run this release
  2. Caching of authorizations in srm and caching of gums (saml-vo-mapping) callouts in gPlazma have been added to version 1.8.
  3. Typo fixed in configuration "linkGroupAuthiorizationFileName" to "SpaceManagerLinkGroupAuthorizationFileName".

beta23 => beta25

  1. added release id into SrmPing
  2. added fast shutdown procedure
  3. improved cog-libraries
  4. Updated FTP client ( used bu pools in srmCopy) to use GFD.47 GETPUT if available.
  5. show AccessLatensy and RetentionPolicy in psu ls -l linkGroup command
  6. better behavior of ftp doors in cafe of client disconnects

beta22 => beta23

  1. FIX OF : srmStatusOfGetRequest returns an inconsistent return code at request level

beta21 => beta22

  1. Fixed PoolManager.conf file corruptions on save in beta21.
  2. Added 'disable ' and 'enabled ' commands in the LoginBroker. This allows to block a door from appearing in 'ls -binary' and thus to drain a door prior to shutting it down (since the srm door will stop generating TURLs for it).
  3. Moved 'rep ls' and 'rep rmclass' processing to background threads.
  4. Modify getFileMetadata so that it does not throw an exception if a user doesn't have proper permission to read its own file. This led to a situation where a user could remove permissions from a file and coudn't modify it any more at all.

beta20 => beta21

beta19 => beta20

  1. added berkeleyDB based pool metadata repository support
  2. added initial version of chimera based cleaner
  3. SRM: Space reservation with retentionpolicy=REPLICA and unspecified accesslatency resolves in reservation of CUSTODIAL-NEARLINE type of space when the access latency optional parameter is not specified.
  4. SRM: An srmPrepareToPut or an srmStatusOfPutRequest returns SRM_NO_FREE_SPACE at file and request level if the space specified is expired instead of returning SRM_FAILURE at file level and SRM_SPACE_LIFETIME_EXPIRED at request level, or (if the space token is no longer valid) SRM_INVALID_REQUEST at request level and SRM_FAILURE at file level.
  5. Remove FTP mover: added custom checksum support
  6. PoolManager: added commands to manipulate Access Latency and retention Policy of a LinkGroup
  7. FtpDoor: Fixed a file descriptor leak in how the socket adapter is used when passive pools are enabled.
  8. FtpDoor: Fixed problem with hanging FTP doors when the transfer times out.

beta18 => beta19

We are currently preparing for the upgrade of your 1.7 system to 1.8. In order to reduce the time it takes to run the upgrade procedure, it turned out to be necessary to change the structure of the StorageInformation files in the pool control directories. ( .../control/SI-<pnfsid> ). But this means that when upgrading from 1.8.0-beta-1.8.0-18 and ealier to 1.8.0-beta-1.8.0-19 and later, it can take awhile before the pools become ready.
What happens ?
When starting a pool after upgrading from 1.8.0-beta-1.8.0-18 to 1.8.0-beta-1.8.0-19, the pools will notice that the StorageInformation files are not compliant to what it expects, so that it tries to recover and gets the information from the PnfsManager. This is what takes some time. So please be patient. You may check the output of the pool logfile while the pool is recovering.
Again : This will allow the 1.8 pools to understand the content of the 1.7 control information and will significantly speed up the upgrade of you production system.
  1. srmGetSpaceTokens returns SRM_INVALID_REQUEST if the space token description provided does not refer to any existing space token
  2. Added support for Space Manager LinkGroup Authorization file

    File syntax
    LinkGroup Name followed by the list of the FQAN, one FQAN on separate line, followed by an empty line, which is used as a record separator, or by the end of file.
    Any line that starts with # is a comment and may appear anywhere.
    File location is specified by defining

    SpaceManagerLinkGroupAuthiorizationFileName = <FILENAME>
    
    in the dCacheSetup

  3. Modified SpaceManager database schema to support mode descriptive properties of the LinkGroups. The LinkGroup definition in the pool manager will change in the next release, which will be available end of this week. These changes should be transparent to the users.
  4. Users can now use FQAN like syntax when specifying roles in the PoolManager.conf file.
    <VO>/<Role=ROLE> instead of 
    <VO>Role=<Role=ROLE> 
    
    The old syntax is still supported.
  5. SRM V1 advisory delete will work with higher reliability.
  6. If the AccessLatency and RetentionPolicy tags are specified in the pnfs directories, the default AL and RP will be based on these tag values.
  7. Space occupied by a file is returned to a Space Reservation to which this file belongs, if the file is deleted.(T0D1)
  8. Option for enabling implicit space reservation for SRM V1.1 and SRM V2.2 is renamed to srmImplicitSpaceManagerEnabled, and is enabled by default (used to be misleadingly called "srmSpaceManager")
  9. These are changes to fix SRM implementation problems compiled by Flavia :

beta15 => beta18

beta14 => beta15

beta13 => beta14

beta11 => beta13

beta10 => beta11

beta9 => beta10

beta8 => beta9

beta7 => beta8

beta6 => beta7

beta3 => beta6




black_bg
Copyright dCache.org © 2003 - 2008