I have a user that today is not replicating the new documents that the others users are adding on my CoucDB database, in other words the user A does not see the docuemnts that the users B, C is adding every day.
I have seen the last document added from the user A to CouchDB and I have seen that the document has a new field
_conflicts: ["2-17d3fcec15fbe3b1eed3e7f8a14eae35"]}
I guess the conflict is in the second revision of the document, Is not it? I have 7 revision about the same document
my question is How I can resolve it? How I can remove this conflict?
CouchDB does not attempt to merge the conflicting revision.
Your application dictates how the merging should be done.
see http://docs.couchdb.org/en/2.0.0/replication/conflicts.html
but generaly speaking the suggested algorithm to fetch a document with conflict resolution:
Get document via GET docid?conflicts=true request;
For each member in the _conflicts array call GET docid?rev=xxx. If any errors occur at this stage, restart from step 1. (There could be a race where someone else has already resolved this conflict and deleted that rev)
Perform application-specific merging
Write _bulk_docs with an update to the first rev and deletes of the other revs.
alos see the version on ruby
Related
I have a notes database and recently we are having an issue across random documents. Users gets error "The linked document cannot be found in the view" when they try to open a document from a view. I am not sure how to reproduce this issue.
I tried the below options to reproduce it, but no luck:
Attaching a document link to one of the Rich text field and deleting the link's target.
Attaching a file as a link and deleting the file in the target.
I am not seeing any pattern with the documents having the issue. The documents accessed few days ago, which was not accessed for more than 2 years also becomes corrupt like this.
Other details:
The Field info tab in Document properties dialog box is empty
There is no scheduled agent running overnight affecting this documents
No change was done to the database recently
We recently migrated to Notes 11.0 from 9.0
Since we have a daily backup of the database, we restore the document from one of the restored copy in which it was working fine.
A guess: did you try re-indexing the view? Shift-F9?
It could also be in the form, or even the view.
What if ...
you open the same document from a different view?
you remove some, many or all fields from the form, do you still see the error?
I think those documents have actually been deleted, and two things make me think that:
You say in a comment that rebuilding the view makes the documents disappear. I'd expect this only if the documents no longer exist.
I have in the past had LotusScript code doing view lookups which produced documents that I knew to be deleted, but the NotesDocument.IsDeleted property was False and the NotesDocument.Items property was Empty (i.e. no fields in the document). This is consistent with your saying the Field info tab is empty.
I'd check if any user action or code could delete documents. You say there are no scheduled agents, but maybe some user action could run code which deletes documents, or some users with the Delete privilege in the database ACL could be directly deleting documents using the Delete key on the keyboard.
Also check the database properties -> Info tab -> User Detail. This should show if there are recent deletions, but the record is probably limited to the last week or less.
I have a field on one of my base page types which I need to update programmatically from an external data feed, so that it can be included in my Smart Search index.
The documents are versioned, but I want to update the published value of this field regardless of checkout state, and obviously avoid any sort of overwrite when documents are checked in.
The field will not appear on the editor form -- or ideally, would conditionally display for Global Admins.
It appears that using the API to update the document without doing a CheckOut fails silently. However if I do a Checkout/Update/CheckIn on a checkout-out page, the author will lose their work I assume?
Any way to handle this "versionless" field via the Kentico data model and API?
I don't think there is a way around updating checked out pages. You can update the page type table directly, but as you mentioned, it will be overwritten when they check in. You could update the version history I believe to make changes to the current data that is checked out, but again, I think that will be lost if the user cancels.
The only way I can think of to solve your issue is to create another table that maps the values you want to the page. Then you don't have to worry about the pages being checked out, you just need to grab the documentID or something. Since the value isn't displayed to the editor, you just have a field that does a lookup on this table.
The preferred and right way is using the API but as you stated, it causes problems if a user has something already checked out and working on it or it's in workflow and not published yet.
If the field you're updating is page type specific, there is one thing specifically I can think of and that's going directly to the database to the page type's database field and perform an update to that field.
Note: this is not recommended unless you know specifically what you're doing and have done full testing on it
The down side of going direct to the database is this will not update the current version since you're using check in/out and workflow. You will also need to update the checked out and current version which means you need to:
Go to the Document itself in the cms_documents table and get the document you are working with.
Then using the fields DocumentCheckedOutVersionHistoryID and DocumentPublishedVersionHistoryID' you can get the version history IDs of the document from theCMS_VersionHistory` table.
Then you can perform an update to the CMS_VersionHistory and your custom page type fields.
You will then need to look in the CMS_WorkflowHistory table and find out if that document is in workflow and in what step.
After you have that workflow history step, use the VersionHistoryID field to go back to the CMS_VersionHistory table and update that record with your data.
Again, not an elegant solution since you are using check in/out and workflow but after some trial and error and testing you should be able to figure it out.
UPDATE
You may also be able to add a custom table or some other linked database table which will allow you to create a global handler. The linked table would be where you perform your updates via API and other calls without versioning or workflow. Then when a user updates a specific page type you could do a check to see when the last time that linked table was updated and update the field(s) you need on update of that particular page (of course by node and document IDs).
Unfortunately you'll have to check it in and out with API. See examples here.
Also you might need to publish it in order to reflect changes on the live site.
I have a strange problem: I want to access documents in a different database (same server). My approach is very close to this one discussed here: http://www-10.lotus.com/ldd/nd85forum.nsf/DateAllFlatWeb/517ef6249d5b9fa6852575cc00503786?OpenDocument
I have only 3 docs in the source database. 2 are created directly, one is copied from another database (these are just test document). We have a generic view that lists thos entries from a view, calcs the links in a form like this:
http://localhost/database.nsf/xpMBK.xsp?action=openDocument&db=dominotest%2Ftest%2Fulcbs%21%21projects%2FFKIE%2FEinsMuB.nsf&view=AMBKEinsAll&documentId=781F14A98A699548C1257C3200316BAC
As you can see we are using an Xpage in the current database and place parameters that point the Xpage to open the document to the source database (notation is server!!database here), a view (this is the one to which I want to return) and finally the unid of the source document.
Now the strange one:
I cannot open the copied document, receiving the
NotesException: Invalid universal id
lotus.domino.local.Database.getDocumentByUNID(Unknown Source)
error.
Even better: if I copy a document that works within the same database (the current one) this document can also not opened anymore!
What's this and can you give me a hint to solve this?
Thanks in advance!
If, by "copied", you mean either manually copied and pasted into the target database or programmatically duplicated via copyToDatabase(), the new copy of the document will be assigned a new UNID; it is not guaranteed to have the same UNID as the original did (and, in my experience, it's rare that it preserves the original). If you're duplicating the document programmatically, be sure to check its new UNID afterward and use that ID in your URL calculation instead.
I've had a problem very similar to this in the past, and the answer turned out to be that I wasn't opening the NSF file that I thought I was opening. I was using NotesDatabase.OpenByReplicaID, and there were two replicas of the database on the server, with different sets of documents. In that situation, Notes gets to pick one of the two replicas -- you have no control over it. The replica that was actually opening contained some of documents corresponding the the UNIDs that I was trying to access, but some of them really were not there and therefore the getDocumentByUNID() method was correct in throwing the "Invalid universal id" error. This was really, really hard to debug.
After I figured it out and removed the second replica from my server, the first thing I did (after testing and confirming the problem went away) was to write an agent that scans a server for duplicate replica IDs.
The UNID:S in a Domino database when it's copied to the database thru copyToDatabase is done like this.
One part of the UNID comes from the database one part is document unique. So if you copy a document from one database to another the document could get the same unid each time. If the unique combination doesn't have a valid document with that combination in the database, the document will get the same UNID everytime. In other cases the document will get a new Id.
More information can be found here
UNID and copytodatabase
Thank you guys for your ideas!
But I was completely wrong #facepalm
The problem was: a colleague coded a bean to access the other database and I didn't noticed that the config document pointed to a replica on another server, so when I copied the document within my database on my local server it was fairly clear that the xpage could not find the copied one - as it resided on the other machine.
Thank you anyway :)
I am trying to update an existing activity stream entry e.g. the title of the entry.
Here I found the code for the creation of the new entry:
Link
But I could not find any reference how do I update an existing entry?
Additional information:
I use IBM Connections 4.5 and the IBM SBT
I create the entries with a system user to other users with the flag
'actionable'
Here my questions:
Which URL?
Which Method (PUT?)?
Which Json?
And another question about the actionable flag:
How can I change the actionable flag for an entry of another user? The description for my own entries is described on slide 37 here: Link
Thank you so much !
Markus
Ok, I think I fully understand the issue now. As suggested this is not supported, but there is a way you can achieve the same result.
First of all why isn't it supported . . .
Events are a point in time (and they were accurate at that point in time)
A new event on the same object supersedes it (as it's now the most interesting) but
doesn't invalidate it (it can still be seen in history)
The Actionable view does not show a rolled up view, instead it shows all events that are marked actionable (and there may be multiple actions related to any given Object)
What you can do . . .
If you want to replace an entry in the Actionable view, you can remove the event from the actionable view (it is just removed from that view and could still be seen in event history)
You can then add another event to the actionable view (which as the latest event will also supersede events in other rolled up views)
Removing the actionable flag is here ->http://www-10.lotus.com/ldd/appdevwiki.nsf/xpDocViewer.xsp?lookupName=IBM+Connections+4.5+API+Documentation#action=openDocument&res_title=Support_for_Saved_and_Actionable_events&content=pdcontent
an activity stream object is treated an an immutable object in IBM Connections.
you can Create Delete and Read.
You can use a rollup-id in IBM Connections.
In order to address the scenario where a user posts a file and 200 people 'like' it, filling up their Activity Stream, rollup needs to be performed. This means:
Only the latest event on any given object is shown
The 2 most recent comments are returned.
http://www-10.lotus.com/ldd/appdevwiki.nsf/xpAPIViewer.xsp?lookupName=IBM+Connections+4.0+API+Documentation#action=openDocument&res_title=Support_for_Rollup&content=apicontent
Just to extend the last answer, delete is not directly supported in the Activity Stream in IBM Connections, though a means of propagating delete based on deletion of an Object was introduced in IBM Connections 4.5.
However it does seem like submitting a new event with an appropriate rollup id is what you're looking for. That way users will see the latest, but the history remains and can be seen if desired.
I am putting a document into couchdb like this:
curl -X PUT http://localhost:5984/bucket/keyname_1 \
-d '{"foo":"bar"}'
This works just fine as long as the document doesn't already exist. If it does exist, I need to include a "_rev" key to validate that I am updating an existing document:
curl -X PUT http://localhost:5984/bucket/keyname_1 \
-d '{"_rev":"2-207df9da","foo":"bar"}'
If the _rev key is missing, or if the _rev key is doesn't match what's in the existing record, this throws a "conflict" error. This is how you know that a revision isn't being thrown away.
But what if I want to write it unconditionally, and bypass this check?
I have a design where I am updating the main record, with the check, and then I want to save a brief summary record with only a couple fields that can be easily grabbed and downloaded for a summary screen. Since I just updated the main record, I know that I have the latest copy. I just want to do a write, and I don't want to have to do a read first just to get the _rev. This would be redundant.
Is there any way that I can modify my request to tell couchbase to skip the _rev check and just put the document in unconditionally?
So you have two documents that you want to update: main and summary. Even if you have the latest version of main, will that actually mean you have the latest version of summary as well or do you just want force the last no matter what? If, then you might have a look at Update handlers: http://wiki.apache.org/couchdb/Document_Update_Handlers
Also, if you go with some kind of secondary GET to fetch the latest _rev, use a cheaper HEAD request.
EDIT: BTW, why having two separate documents? Is the Summary one in fully complete with the Main one? All all the members found there? Could you then perhaps use Show or list functions? http://wiki.apache.org/couchdb/Formatting_with_Show_and_List
Instead of PUTting the record, why not POST a new one?
That way, you keep a full record of all such log events, but you don't have to deal with document conflicts.