daitss / core

DAITSS: Dark Archive In The Sunshine State
GNU General Public License v3.0
9 stars 2 forks source link

Disseminated package errors in the xml-resolution stage #801

Closed szanati closed 6 years ago

szanati commented 6 years ago

I got the following error while a UF package, USACH00236_00001 ieid EOI7XHCH7_GWFQU1, was disseminating:

Must specify a string with one or more characters

/opt/web-services/sites/core/shared/bundle/ruby/1.9.1/gems/libxml-ruby-2.7.0/lib/libxml/parser.rb:84:in string' /opt/web-services/sites/core/shared/bundle/ruby/1.9.1/gems/libxml-ruby-2.7.0/lib/libxml/parser.rb:84:instring' /opt/web-services/sites/core/shared/bundle/ruby/1.9.1/gems/libxml-ruby-2.7.0/lib/libxml/document.rb:76:in string' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/metadata.rb:40:inblock in old_events' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/metadata.rb:40:in map' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/metadata.rb:40:inold_events' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/metadata.rb:18:in next_event_index' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/wip/disseminate.rb:18:inblock in disseminate' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/wip/journal.rb:16:in step' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/wip/disseminate.rb:17:indisseminate' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/wip/process.rb:82:in block in spawn' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/wip/process.rb:66:infork' /opt/web-services/sites/core/releases/20170924000001/lib/daitss/proc/wip/process.rb:66:in spawn' /opt/web-services/sites/core/current/bin/pulse:161:inblock in start_wips' /opt/web-services/sites/core/current/bin/pulse:158:in each' /opt/web-services/sites/core/current/bin/pulse:158:instart_wips' /opt/web-services/sites/core/current/bin/pulse:194:in block in ' /opt/web-services/sites/core/current/bin/pulse:192:inloop' /opt/web-services/sites/core/current/bin/pulse:192:in `'

cchou commented 6 years ago

It looks to be related to old digi-prov or namespace issue. Will need to hack my local database to reproduce the dissemination error for further troubleshoot.

Stephen,can you put together an original SIP and put it on daitss-test for me? Please also put the AIP descriptor before dissemination on daitss-test. Thank you, -Carol

szanati commented 6 years ago

Carol, I put on the daitss-test site 3 different packages. The first is a package called USACH00236_00001.tar that I recreated from the original files . The second package is EOI7XHCH7_GWFQU1.tar and I curled it from tarchive. Its the original package that was submitted back in 2013. The last package is EOI7XHCH7_GWFQU1_Stash.tar and it is the package from the current error found in /var/daitss/data/stash/Github_801. I added it in case it is helpful.

szanati commented 6 years ago

We got another package that was disseminating when It errored with the same error: Must specify a string with one or more characters. Package UF00026177_00001 and ieid E20100128_AAAAHS. The package is too big to put on the daitss-test site but it has the same error as the other package.

szanati commented 6 years ago

We have received yet another package that was disseminating that errored with the same error: Must specify a string with one or more characters. Package CFE0003386 ieid EAOOUGQGD_4D79WW. Could this error be related to the code roll out on Monday?

cchou commented 6 years ago

Possibly, since one of the rollout is to bring in the old-digiprov in the old premis namespace. I did some dissemination testings with the rollout code, but none has this error, so it most likely relates to some existing packages on darchive that's why I need to hack it to reproduce the error.

We could undo the issue #797 rollout for now until a fix for this one is done.

szanati commented 6 years ago

I think that would be best since 3 disseminations so far have been affected by it.

cchou commented 6 years ago

Well, rolling back means the old digi-prov in those packages will not get carried over when they get disseminated, so I think we just need to move forward. Testing and fixing dissemination on old packages are always tricky since there are so many variations of packages in FDA archive, especially now we have to deal with multiple versions of premis namespace, and they are not easy to reproduce and troubleshoot.

cchou commented 6 years ago

I have a fix, it has been tested on ripple, https://core.ripple.fcla.edu/package/EZ475RCDQ_1QT7E4.

Please let me know when you would like it to roll out to prod.

szanati commented 6 years ago

Carol,

I think you should roll out to prod as soon as you can before we get more dissemaintions being affected. Just let me know we you can do it.

cchou commented 6 years ago

I can roll them out to prod either at 5:30pm today or 5pm tomorrow. Let me know which one work for you.

szanati commented 6 years ago

5pm tomorrow will work best for me. So I will stop daitss just before 5pm tomorrow and email you that it is ready for you to roll out.

szanati commented 6 years ago

Code was rolled out on 10-3-17 and all 3 packaged archived. I will now close this issue.