You are viewing an old version of this page. View the current version.

Compare with Current View Page History

Version 1 Next »

The outgoing programmer is responsible for all of the end-of-expedition (EOX) activities.

The goal is to ensure a good copy of the data gets back to HQ.

Please read this entire document and EOX - detail. Expect to take 3 to 4 hours to complete this exercise. (I keep saying this only takes 15 minutes. But that really is just the individual database dump processes.) Do not hurry. If you are "fried"--work together. 

0. Readiness assessment.


Confirm with science party and technical staff before proceeding.

  • Are we done collecting data for this laboratory area?
  • Has all the data content that you want to go home been:
      uploaded to the database
      copied to data1 OR
      copied to the EXPEDITION PROCEEDINGS volume
    ?
  • Where is the data? A data disposition form is available in s:\data1.
    It is helpful to the data librarian to have that filled out.

If no, obtain estimates when the data collection will be done. Schedule accordingly.   

1. Notify everyone. Database snapshot in progress.

  • Notify jr_developer and the shipboard technical staff that the EOX database backup is in progress.
  • The database is available for reports, but content cannot be changed while this activity is in progress.

2. Make the database read only.


Connect to the BUILD box. Execute all of the following steps from that system.


Run PuTTY.
Connect to oracle@k1.ship.iodp.tamu.edu
From the prompt run this command, supply the oracle user's credentials for k1

. oraenv
ORACLE_SID = [LIMSJR] ? LIMSJR
sqlplus your-name_dba
sql>
alter tablespace labware read only;


At the SQL> prompt, run the last command.
Stop. Leave this window open for use below.

Technical note. This command prevents write access to any table content that resides in the storage facility under the "labware" table space. In our database implementation, this happens to be all tables owned by the LIMS schema. Use of Drill Report ("ops" schema) is not affected. Aspects of DescLogik software configuration are also still open to maintenance, though use of DescLogik itself is locked out by this change.

3. Full snapshot as oracle@k1.


Run another PuTTY session.
Connect to oracle@k1.ship.iodp.tamu.edu
From the prompt run these commands--change the expedition and yourname  (will require logging in as a DBA User see PWSAFE) - - - JON Put Your Password in "QUOTES"

NOTE: While performing the EOX for expedition 384, it was discovered that the path represented by "data_pump_dir" did not exist.  You can discover what the data_pump_dir is supposed to be using this query (logged in as your DBA account):

select * from dba_directories;

The results have a lot of stuff that's not useful, but the DATA_PUMP_DIR should be there.  Currently it is /backup/LIMSJR/dpdump.  Note that "backup" is singular (it used to be plural) and there is apparently some desire to change this, so keep that in mind.  If you find that the folder does not exist, the command in this step will fail and you may need to get an MCS or a DBA on shore to help create it.

You can also try:

create directory data_pump_dir as '/backup/export';
grant read, write on directory data_pump_dir to transfer, system;

10/5/2021: Folder did not exist, as above, but MCS suggested I try creating it myself (in Putty, logged in as Oracle) using mkdir commands.  That initially seemed to work, but it broke some kind of symbolic or virtual link that caused the file to be written to the wrong physical location (with reduced disk space), so not recommended.

. oraenv
ORACLE_SID = [LIMSJR] ? LIMSJR
# Note that the folder you're cd-ing to here should be the "data_pump_dir" referenced in the next command.  This comes from the dba_directories table (see note above).
cd /backup/LIMSJR/dpdump
time expdp YOUR_DBA_ACCOUNT directory=dmpdir full=y logfile=limsjr_full-YOUR_EXP_NUMBER-export.log dumpfile=limsjr_full-YOUR_EXP_NUMBER-YOUR_INITIALS.dmpdp


Statistics. Typical durations for the last step from previous expeditions. 344S: 16 min 23 sec. 345: 13 min 54 sec. 346: 18 min 46 sec. 341S: 11 min 42 sec. 351: 12 min 9 sec; 360: 16 min 57 sec; 363: 3 min 21 sec; 367: 3 min 38 sec.

ODA for 3 expeditions of data (11.1 GiB): 4 min 17 sec.
396: 6 min 54 sec

4. Restore database read/write capability.

From the prompt [oracle@k1 ~]$

run these command, supply the appropriate credential

sqlplus YOUR_DBA_ACCOUNT
sql>
alter tablespace labware read write;
exit

Stop. Leave this window open for use below.

5. Compress the exports for transport. Cleanup. 

From the prompt [oracle@k1 ~]$
run these commands

cd /backup/LIMSJR/dpdump
time bzip2 --best limsjr*YOUR_EXP_NUMBER*.dmpdp
# once the above completes, then run this
ls -l
md5sum *.bz2

Cut and paste the digest keys and file listing into a notification email. Enables remote users to verify the integrity of the files.

           Please log these in a txt file in the \AD\support\'expedition named sub directory'\ see other expeditions for example.

Statistics. 367: 22 min to compress both files (~5 GiB to < 1 GiB); 345: 22 min 53 sec for bzip step. 351: full 137 min 46 for ~8 GiB content.  360: 67 min 27 sec for ~5 GiB file. 45-60 min for high recovery expeditions.

Exp 396: we had trouble initially because we ran out of disk space.  This was apparently due to my creating the /backup/LIMSJR/dpdump folder under the Oracle account.  This broke some kind of link-up that gives the backup folder lots of additional drive space.  Don't do that.

Need to test DMPDP.
Need to post log results.

6. Notify everyone. Database snapshot process is complete.

 Notify jr_developer and expedition technical staff that the process is done.

Speak with the MCS. They will pick up the data from here.

oemjr:/backup/export
    limsjr*.dmpdp
    limsjr*.log


The MCS pick up the full export from  oemjr:/backup/export.

Inquire with the MCS. Ensure the (above) database content and the (below) ASMAN content are being taken to media for transport to HQ.

  • No labels