Bug #6800
SOLR indexes malformed strings - identifier, id
100%
Description
During maintenance of mnTestLTER content on cn-sandbox-ucsb-1.test.dataone.org, it was found that results from a SOLR query returned records that contained malformed "identifier" and "id" strings. There are a total of 18 of such records.
Query:
curl -s -X GET "https://cn-sandbox-ucsb-1.test.dataone.org/cn/v1/query/solr/?q=datasource:urn\:node\:mnTestLTER"
List of malformed identifiers:
ld>:10.6073/pasta/e2c4d7746fc2dcda6bf62cba99389567¶
>oi:10.6073/pasta/18e14273378ba371de1833a41d32301d¶
otrends/2927/2/d5f755665edc9ckage/data/eml/ecotrends/2927/2/d5f755665edc967b5c9195366cbb6c8d¶
417/2.lternet.edu/package/report/eml/ecotrends/10417/2¶
rends/8468/2t.edu/package/report/eml/ecotrends/8468/2¶
85a238d13910aeaf02db6aage/data/eml/ecotrends/3920/2/ea7fd9dfab85a238d13910aeaf02db6a¶
bff04f425a9409e595b9ckage/data/eml/knb-lter-and/2722/6/924fa927309abff04f425a9409e595b9¶
071ce2bca9a8f73a007241b96f2240f1ml/ecotrends/1671/2/071ce2bca9a8f73a007241b96f2240f1¶
age/data/eml/.lternet.edu/package/data/eml/ecotrends/10568/2/a115d2a94070ca2085996903277db48a¶
34fd95f1581a290a6a35e1ege/data/eml/ecotrends/8459/2/5644b03fb34fd95f1581a290a6a35e1e¶
ld>ps://pasta.lternet.edu/package/data/eml/ecotrends/4473/2/f0e51aae0a51606233df892959d3a56d¶
otrends/8606/2/f0b80ea995d9fckage/data/eml/ecotrends/8606/2/f0b80ea995d9f6d585a2c9b2d99a7872¶
age/report/em.lternet.edu/package/report/eml/knb-lter-arc/10272/3¶
11085/2/18e3871ec8c06eb1420867cd2c6ata/eml/ecotrends/11085/2/18e3871ec8c06eb1420867cd2c6a0cb6¶
46a31afb6bfbeb683a12e938e/data/eml/ecotrends/6868/2/c2ca91ad46a31afb6bfbeb683a12e938¶
f3e297e1dd</pasta/a76c718448bff15dc8b004f3e297e1dd¶
211835310dbf305f731b9cce54b41bdeml/ecotrends/6584/2/e211835310dbf305f731b9cce54b41bd¶
d53e0930ernet.edu/package/data/eml/ecotrends/1586/2/c21358a57bffd9e176f634c1d53e0930¶
It is as if the string is being overwritten by left-over string-buffer content - for example, the malformed identifier "d53e0930ernet.edu/package/data/eml/ecotrends/1586/2/c21358a57bffd9e176f634c1d53e0930" contains the wrong sub-string "d53e0930" in position 0 through 15, thereby replacing the correct sub-string "https://pasta.lt". The same type of malformed strings is also found on cn-stage-ucsb-1.test.dataone.org.
The full output from the SOLR query may be found in the attached text document "cn-sandbox-ucsb-1-SOLR.txt".
Related issues
History
#1 Updated by Matthew Jones about 9 years ago
The MDC project also found similarly malformed strings in the production SOLR index. Peter Slaughter is investigating.
#2 Updated by Skye Roseboom about 9 years ago
- translation missing: en.field_release set to 2
#3 Updated by Rob Nahf about 9 years ago
- Related to Bug #6870: Fix handling of identifiers with url-escaped characters added
#4 Updated by Lauren Walker about 9 years ago
Here is an identifier that is experiencing this bug right now, on sandbox 2:
_NA_v2.0.xmlMAP_PRESENTVEG_C3Grass_RelaFrac_NA_v2.0.xml
#5 Updated by Skye Roseboom about 9 years ago
- Status changed from New to In Progress
- % Done changed from 0 to 30
#6 Updated by Skye Roseboom about 9 years ago
- Target version set to CCI-1.5.1
#7 Updated by Skye Roseboom about 9 years ago
- Related to deleted (Bug #6870: Fix handling of identifiers with url-escaped characters)
#8 Updated by Skye Roseboom about 9 years ago
- Duplicated by Bug #6870: Fix handling of identifiers with url-escaped characters added
#9 Updated by Skye Roseboom almost 9 years ago
- Target version deleted (
CCI-1.5.1)
#10 Updated by Skye Roseboom almost 9 years ago
- Subject changed from SOLR indexes malformed identifier and id strings for mnTestLTER to SOLR indexes malformed strings - identifier, id
Problem effects all environments. It also effects more than the identifier and id fields. Problem has been observed in date fields, full-text field, origin, author, fileID.
Problem is not consistent - in that re-indexing a solr record wit mangled values creates a new solr record without any mangled values.
Debugging indicates that the index-processor is sending proper solr xml messages to the solr server, but the string values are being misinterpreted by the solr server. This is supported by output of the xml payload being generated at d1's index processor and the xml payload received at the solr server - both of which when logged appear to be proper - no mangled string values.
Problem persists and will take more work to discover cause.
Working on some scripts to help detect and clean up solr records and re-index documents which get mangled.
#11 Updated by Peter Slaughter almost 9 years ago
- File pidsInSolrNotInObjectStore.txt
added
Ran a test on 6/26/2015 on cn-ucsb-1.dataone.org that fetched all pids from solr and then checked in object store (in this case guid field from 'systemmetadata' table). The attached
file 'pidsInSolrNotInObjectStore.txt' shows the pids that were in solr but not in the systemmetadata table.
#12 Updated by Matthew Jones over 8 years ago
This problem also extends to the AuthoritativeMN field. Here's an example of it being mangled:
evolve whRYAD
which can be seen here:
https://cn.dataone.org/cn/v1/query/solr/?fl=identifier,authoritativeMN&q=id:*dryad.gp23s/1%3Fver*
#13 Updated by Skye Roseboom over 8 years ago
Issue is a cause of running solr3.x on a Java7 JDK. Solr3.x line is no longer patched by solr community so issue will not be resolved until deployment of solr5 with the CCI V2.0
#14 Updated by Skye Roseboom over 8 years ago
- % Done changed from 30 to 50
- Status changed from In Progress to Testing
New version of solr is being deployed to production and search index is rebuilding. No mangled strings have been detected in stage or production.
#15 Updated by Skye Roseboom over 8 years ago
- Status changed from Testing to Closed
- % Done changed from 50 to 100
Solr 5 installed into production and index rebuilt. Problem appears to have gone away with solr5
