1. Reminder: what is Atomic Update?
Once you have indexed the content you need in your Solr index, you will want to start thinking about your strategy for dealing with changes to those documents. Solr supports three approaches to updating documents that have only partially changed.
The first is Atomic Update. This approach allows changing only one or more fields of a document without having to reindex the entire document.
The second approach is known as in-place updates. This approach is similar to atomic updates (is a subset of atomic updates in some sense), but can be used only for updating single valued non-indexed and non-stored docValue-based numeric fields.
The third approach is known as optimistic concurrency or optimistic locking. It is a feature of many NoSQL databases, and allows conditional updating a document based on its version. This approach includes semantics and rules for how to deal with version matches or mis-matches.
Atomic Updates (and in-place updates) and Optimistic Concurrency may be used as independent strategies for managing changes to documents, or they may be combined: you can use optimistic concurrency to conditionally apply an atomic update.
1.2. Choice in Datafari
Doing partial document updates like this can be particularly useful in Datafari when we need to add a type of metadata to documents that takes a long time to extract. First we index documents by extracting the “simple” metadata, so that these documents are quickly available for search. Then we can extract the more complex metadata to complete documents information and improve results accuracy.
Given that in Datafari all the fields we wish to update are indexed and stored, we opted for the Atomic Update approach.
We imposed ourselves not to have incomplete documents, in a sense that an atomic update should only update solr documents that have been created by the “standard” crawl (otherwise, for instance in case a full crawl creates a solr document, then an incremental standard crawl deletes this document, before the atomic update comes in while still having the document in its intermediary solr, it would create a new solr document with incomplete data). For that, optimistic concurrency in Solr gives us the ability to only update documents already created.
2. Processing principle
Without Atomic Update, if we want to enrich a Solr collection with OCR data using the current Annotator functionality, we create a dedicated OCR job targeting the same collection. This job will overwrite the collection with its data. The OCR job must be up to date to guarantee fresh documents content. If we want to add Spacy data to the same collection (a Spacy server allows us to extract natural language entities using Hugging Face AI models), we have to add it to the the same job as the OCR one. Indeed, if we were to create a dedicated Spacy job, it would completely overwrite the collection with its data; and since it has no OCR related data, the OCR related field(s) would be emptied (and vice versa when the OCR job passes by). This job being sequential, it will only be as fast as the slowest computation step (either OCR or Spacy in our case). We cannot parallelize the computations.
We are therefore very careful when designing the job sequencing if we don't want the fresh content of documents to be overwritten by the Spacy or OCR job.
Using Datafari’s Atomic Update Service solves this problem. We can update documents with job-specific fields only. The operating principle is somewhat similar to proces sing without Atomic Update. Compared to our current Annotator functionality, we can now do it in several separate jobs. In the scenario mentioned earlier, we would have one fast job, to index the collection to be enriched, one job for the OCR and one for Spacy. Compared to the previous mechanism, the OCR and Spacy jobs must index the documents in separate collections (see the illustration below). Then, the Atomic Update Service can be called. It will retrieve the Spacy and OCR specific fields to update the global collection.
3. Run Atomic Update Service
The Atomic Update Service is available as an executable file with a configuration file used to set up the update jobs. You will need to launch one Atomic Update job per collection used to update the destination collection. Using the same exemple as above, you will launch one Atomic Update job for the Spacy collection and another for the OCR collection.
Find this component in Datafari in [DATAFARI_HOME]/bin/atomicupdates containing these files:
atomic-updates-launcher.sh: to run Atomic Update (more details below)
atomicUpdate-cfg.json: to configure your job(s) (examples of values supplied)
atomicUpdate-example.json: the configuration file explained
atomicUpdate-log4j2.xml: the log configuration file (A user does not need to modify this file)
datafari-solr-atomic-update-… .jar : the executable file
atomicUpdateLastExec: contains the job(s) status and last execution date and time (more details in ).
To modify these files you need to be datafari user or belong to root group.
3.1. Configuration
For each Atomic Update job, you must configure the following:
the collection used to update the target collection, that is the location and name of the Spacy collection, for exemple.
the target collection
the source collection fields used to update the destination collection fields
(Optional) the mapping between source fields and destination fields, when they are not the same.
The configuration file is “atomicUpdate-cfg.json” and here is the way to set it up :
{
"logConfigFile": "", // Specify the log configuration file location if it is different from the provided one in the "atomicupdates"
directory or if you want to move it in another location.
"jobs": {
"JOB_1": { // Put the name you want for this configuration
"source": { // Define the Intermediary Solr, OCR or Spacy in our example above
"baseUrl": Solr Cloud base Url for the source Collection used to update target Collection.
You can specify Solr or Zookeeper host, but prefer the Zookeeper host as Datafari use it to dispatch to all Solr hosts you have.
(For information) The syntax for Solr host is: "http://datafari_domain:8983/solr", ex: "http(s)://localhost:8983/solr" ; you need to specify all Solr hosts.
The syntax for Zookeeper is: "datafari_domain:2181", ex: "localhost:2181" ; No http prefix because it's another protocol.
Whatever host type, you can define several severs by separating URLs with comma, but using Zookeeper, there is only one server. Example with solr host: "http://solr1:8983/solr, http://solr2:8983/solr,...".
"solrCollection": the Solr source Collection for JOB_1. Example "Spacy".
},
"destination": { // Define the final Solr, FileShare in our example above.
"baseUrl": Solr Cloud base Url for the target Collection. The syntax is the same as in "source" block.
"solrCollection": the Solr target Collection for JOB_1. Example "FileShare".
},
"fieldsOperation": { // the fields of the source collection and Atomic Update operation like: set, add, remove, etc...
// the "set" operation will be the more appropriate value for most cases, as it replaces the target value with the source value.
// see more about operations available here: https://solr.apache.org/guide/solr/9_5/indexing-guide/partial-document-updates.html#atomic-updates
"field_1": "set",
"field_2": "add",
"field_3": "add-distinct",
"field_4": "set"
},
"nbDocsPerBatch": The documents are selected and updated per batches.
This represents the number of documents per batch fetched from the intermediary Solr ("solrCollection" in the "source" parameter, for instance the Spacy Solr collection)
up to the final Solr ("solrCollection" in the "destination" parameter, the FileShare collection in our illustration above).
Each batch is stored in RAM so this number depends on the data size retrieved (i.e fields and their content).
You can give a try for instance to values 1000 for OCR sources and 2000 for Spacy sources, and play with them to optimise your atomic updates performances.
"fieldsMapping": { // Optional: to specify a mapping between source and destination collections (remove the content of the bloc if you don't need mapping, like this: "fieldsMapping": {})
"field_3": "dest_field_1",
"field_2": "dest_field_2"
}
},
"JOB_2": { // Put the name you want for this configuration
...
}
}
}
The atomicUpdate-example.json file pretty much repeats what has been said here.
Exemple:
{
"logConfigFile": "/opt/datafari/tomcat/conf/atomicUpdate-log4j2.xml",
"jobs": {
"OCR": {
"source": {
"baseUrl": "dev.datafari.com:2181",
"solrCollection": "OCR"
},
"destination": {
"baseUrl": "dev.datafari.com:2181",
"solrCollection": "FileShare"
},
"fieldsOperation": {
"previewContent": "set",
},
"nbDocsPerBatch": 1000,
},
"SPACY": {
"source": {
"baseUrl": "dev.datafari.com:2181",
"solrCollection": "Spacy"
},
"destination": {
"baseUrl": "dev.datafari.com:2181",
"solrCollection": "FileShare"
},
"nbDocsPerBatch": 2000,
"fieldsOperation": {
"entity_product": "set",
"entity_loc": "set",
"last_author": "add-distinct"
},
"fieldsMapping": {
"last_author": "author"
}
}
}
}
If there is a syntax error in the configuration file, the job will not start and no log is generated. Otherwise, at least one line is entered to indicate that the job has been started.
If there are configuration data errors, the job fails and you can check errors in the log file (see more in https://datafari.atlassian.net/wiki/spaces/DATAFARI/pages/2939387906/Atomic+Update+Management#4.3.-Logs).
3.2. Run
You can manually launch a job with the command (no need for a specific permission): bash atomic-updates-launcher.sh <job_name> [fromDate]
<job_name> one job to be specified and refers to the name (case sensitive) put in the configuration file “atomicUpdate-cfg.json”:
"jobs": {
"OCR": {
...
},
"SPACY": {
...
}
}
[fromDate] (optional) you can force the date from which to select documents (based on last_modified Solr field). The documents we speak about are those of the intermediate Solr of the job you want to run. The expected date format is either "yyyy-MM-dd HH:mm" or "yyyy-MM-dd".
It may be convenient to force this date if you want to update documents from a date before the last execution date of the job (which is the default behavior). An other good reason to force this date, is to avoid a full crawl after a previous crash of the job. In this particular case, you will need to modify the job status to DONE in the “atomicUpdateLastExec“ file, otherwise the full crawl will still be executed. To clarify more, see https://datafari.atlassian.net/wiki/spaces/DATAFARI/pages/2939387906/Atomic+Update+Management#When-the-job-is-done and https://datafari.atlassian.net/wiki/spaces/DATAFARI/pages/2939387906/Atomic+Update+Management#What-happens-when-a-job-fails%3F.
Specify "full" (not case sensitive) to force full crawl. A Full crawl may be necessary if for some reason, one or more of your jobs in the Atomic Update chain has run a full crawl causing the overwriting of fields to be updated by Atomic Update.
When the job is launched without “fromDate” specified, if it is the first run, a full crawl is done. The next runs will select documents from the last execution time of the job (start time).
Exemple:
bash atomic-updates-launcher.sh SPACY or bash atomic-updates-launcher.sh SPACY full or bash atomic-updates-launcher.sh SPACY "2023/02/05"
It is possible to run several Atomic Update jobs at the same time, for example the OCR and Spacy jobs, given that a priori, these jobs will not update the same Solr fields.
To create a scheduled Service use the Cron tool. Enter the command:
crontab -e
Then add the line (replace [DATAFARI_HOME] by your Datafari’s location):
15 1 * * * [DATAFARI_HOME]/bin/atomicupdates/atomic-updates-launcher.sh <job_name> │ │ │ │ │ │ │ │ │ │ │ │ │ │ |_________ Day of Week (0 – 6) (0 is Sunday, or use names) │ │ │ |____________ Month (1 – 12),* means every month │ │ |______________ Day of Month (1 – 31),* means every day │ |________________ Hour (0 – 23),* means every hour |___________________ Minute (0 – 59), * means every minute
In this example, the Atomic Update runs everyday at 1:15 AM.
Here is another example for a weekly job, each Monday at 4 AM:
0 4 * * 1 /opt/datafari/bin/atomicupdates/atomic-updates-launcher.sh OCR
When the job is done
At the end of the job, its status and execution date is written in the file “atomicUpdateLastExec” (in [DATAFARI_HOME]/bin/atomicupdates).This file is created after the first execution of Atomic Update and is common to all jobs configured. Example, for the OCR
# #Tue Mar 13 15:51:41 CET 2024 OCR.STATUS=DONE (FAILED if so) OCR.LAST_EXEC=2024-03-12T14\:00\:00.428Z SPACY.STATUS=DONE SPACY.LAST_EXEC=2024-03-13T15\:30\:00.523Z
What happens when a job fails?
When a job fails, the default behavior is to run a full crawl next time. As soon as the status FAILED is mentioned in the “atomicUpdateLastExec“ file, full crawl is forced. We made this choice thinking of a scheduled job that might fail for a temporary reason. Forcing a full crawl at next run, we give a scheduled job a chance to turn back to a safe state by itself with the guaranty of data integrity.
A temporary reason may come with a system operator who has restarted the Solr server of the final collection, for instance. If the Atomic Update job was running, it will lose access to the Solr collection to update then the job will end with a FAILED status.
If you don't think it's necessary to do a full crawl, you can change this behavior by modifying the job status to DONE in the “atomicUpdateLastExec“ file and running it manually, specifying from which document date you want to resume (“fromDate” parameter explained above).
3.3. Logs
The log file is [DATAFARI_HOME]/logs/atomic-update.log.
On each line of the file, the job name is specified, as this log file is common to all Atomic Update jobs.
It contains at least one line specifying that the job has started.
If everything is going fine, there are lines with the number of documents processed for each successful batch of documents updated.
If no error occurs during the run, the final line specify the job status as "DONE". Otherwise, the final line specifies “FAILED” status and the total number of documents processed appears at previous line.
INFO 2024-03-12T15:17:48Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Job started ! INFO 2024-03-12T15:17:48Z (main) - Datafari-jar|Atomic Update|datafari.save.JobSaver|The Last execution file doesn't exists yet. It will be created by the first Atomic Update Job execution. INFO 2024-03-12T15:17:48Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Job: Select documents modified from: null (null indicates a full crawl)) INFO 2024-03-12T15:17:48Z (main) - Datafari-jar|Atomic Update|datafari.save.JobSaver|Job OCR RUNNING INFO 2024-03-12T15:17:49Z (main) - Datafari-jar|Atomic Update|solrj.impl.ZkClientClusterStateProvider|Cluster at localhost:2181 ready INFO 2024-03-12T15:17:49Z (main) - Datafari-jar|Atomic Update|solrj.impl.ZkClientClusterStateProvider|Cluster at localhost:2181 ready INFO 2024-03-12T15:17:50Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Docs processed : 1000 INFO 2024-03-12T15:17:53Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Docs processed : 2000 INFO 2024-03-12T15:17:56Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Docs processed : 3000 INFO 2024-03-12T15:17:59Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Docs processed : 3900 INFO 2024-03-12T15:17:59Z (main) - Datafari-jar|Atomic Update|datafari.save.JobSaver|Job OCR DONE
Example of first run and failed job:
INFO 2024-03-12T14:58:24Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Job started ! INFO 2024-03-12T14:58:24Z (main) - Datafari-jar|Atomic Update|datafari.save.JobSaver|The Last execution file doesn't exists yet. It will be created by the first Atomic Update Job execution. INFO 2024-03-12T14:58:24Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Job: Select documents modified from: null (null indicates a full crawl)) INFO 2024-03-12T14:58:24Z (main) - Datafari-jar|Atomic Update|datafari.save.JobSaver|Job OCR RUNNING INFO 2024-03-12T14:58:24Z (main) - Datafari-jar|Atomic Update|solrj.impl.ZkClientClusterStateProvider|Cluster at localhost:2181 ready ERROR 2024-03-12T14:58:25Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Job: Total number of documents processed: 0 java.lang.NullPointerException: null at com.francelabs.datafari.solraccessors.DocumentsUpdator.createSolrDocToUpdate(DocumentsUpdator.java:114) ~[datafari-solr-atomic-update-6.1-dev-Community.jar:?] at com.francelabs.datafari.solraccessors.DocumentsUpdator.updateDocuments(DocumentsUpdator.java:62) ~[datafari-solr-atomic-update-6.1-dev-Community.jar:?] at com.francelabs.datafari.SolrAtomicUpdateLauncher.main(SolrAtomicUpdateLauncher.java:133) [datafari-solr-atomic-update-6.1-dev-Community.jar:?] INFO 2024-03-12T14:58:25Z (main) - Datafari-jar|Atomic Update|datafari.save.JobSaver|Job OCR FAILED
Here is the beginning lines of log at next run after job failed:
INFO 2024-03-12T17:06:59Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Job started ! INFO 2024-03-12T17:06:59Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Job: Select documents modified from: 2024-03-12T14:28:32.428Z (null indicates a full crawl)) INFO 2024-03-12T17:06:59Z (main) - Datafari-jar|Atomic Update|francelabs.datafari.SolrAtomicUpdateLauncher|OCR Job: Last state was FAILED, so a full crawl is done for this run. INFO 2024-03-12T17:06:59Z (main) - Datafari-jar|Atomic Update|datafari.save.JobSaver|Job OCR RUNNING and so on...