RavenDB 3.0, Stale indexes upon start of RavenDB 3457.

[Derek den Haas]

Upon boot of RavenDB, after a normal shutdown, and all indexes weren't stale, this shows in the RavenDB log:

Could not open index SupplierAll. Recovery operation failed, forcibly resetting

index Index directory 'C:\Users\Derek\Desktop\RavenDB\Server\Database\Databases/

EasyFlor-Dev\Indexes\135' does not exists for 'SupplierAll' index.

Could not open index SupplierCategoryAll. Recovery operation failed, forcibly re

setting index Index directory 'C:\Users\Derek\Desktop\RavenDB\Server\Database\Da

tabases/EasyFlor-Dev\Indexes\121' does not exists for 'SupplierCategoryAll' inde

x.

Could not open index SupplierContactAddressAll. Recovery operation failed, forci

bly resetting index Index directory 'C:\Users\Derek\Desktop\RavenDB\Server\Datab

ase\Databases/EasyFlor-Dev\Indexes\136' does not exists for 'SupplierContactAddr

essAll' index.

Not all indexes are stale at the moment, but it's now deleting other indexes:

 

IndexDeleteOperation

 

4 minutes ago (09/01/14, 7:21am)

 

Running

 

CompanyAll

I don't see any error's though, and the indexing is pretty fast, only 5 minutes of "stale" indexes for 194.000 documents.

Running pretty good afterwards (zero other error's), congrats with that! It also feels pretty robust!

P.s. the indexes will get removed after every "reboot" (as in q in console, then start ravendb by clicking Raven.Server.exe), so not only the first "reboot". If you want I can share the Database.

[Oren Eini (Ayende Rahien)]

Yes, we had an issue with order of disposal that meant that were forgot that we were indexing the last batch of items.

Fixed now.

Oren Eini

CEO

Mobile: + 972-52-548-6969

Office:  + 972-4-622-7811

Fax:      + 972-153-4622-7811

--
You received this message because you are subscribed to the Google Groups "RavenDB - 2nd generation document database" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ravendb+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Oren Eini (Ayende Rahien)]

Please test the next build as well

[Derek den Haas]

Could not open index BindFeatureAll. Recovery operation failed, forcibly resetti

ng index Index directory 'C:\Users\Derek\Desktop\RavenDB\Server\Database\Databas

es/EasyFlor\Indexes\16' does not exists for 'BindFeatureAll' index.

Could not open index EntityAttachmentMappingAll. Recovery operation failed, forc

ibly resetting index Index directory 'C:\Users\Derek\Desktop\RavenDB\Server\Data

base\Databases/EasyFlor\Indexes\15' does not exists for 'EntityAttachmentMapping

All' index.

Will appear every time I start RavenDB, this is from my smaller database with only 5 indexes (2 fail all the time). The only thing I didn't see was the message "Database upgrading" upon entering that database, only on the larger database (which has the same indexes (and 100 more) but is fully working).

The indexes are pretty simple:

EntityAttachmentMappingAll:

from result in docs.EntityAttachmentMappings

select new {

Id = result.__document_id,

URL = result.URL

}

Op maandag 1 september 2014 07:51:48 UTC+2 schreef Oren Eini:

[Derek den Haas]

P.s. Main.js bug:

                for (var i = 0; i < rowProperties.length; i++) {

                    var prop = rowProperties[i];

                    var cellValue = rowData[prop];

                    // pass json object when not custom template!

                    if (typeof cellValue === "object" && this.getCellTemplateName(prop, rowData) !== cell.customTemplate) {

                        cellValue = JSON.stringify(cellValue, null, 4) || "";

                    }

                    if (cellValue.length > 250) {

                        cellValue = cellValue.substring(0, 250);

                    }

                    this.addOrUpdateCellMap(prop, cellValue);

                }

rowData contains: { Color: null }

cellValue is now NULL and when entering if (cellValue.length > ... it will fail, because you can't access a NULL value.

Op maandag 1 september 2014 08:10:32 UTC+2 schreef Derek den Haas:

[Derek den Haas]

P.s. 2: function: row.prototype.fillCells = function (rowData) {

Op maandag 1 september 2014 08:21:10 UTC+2 schreef Derek den Haas:

[Oren Eini (Ayende Rahien)]

Huh? I don't follow.

[Derek den Haas]

I'm sorry Oren,

I've updated to the newer version, and on 1 database the index upon restart will still be reset (2 indexes out of 5).

On another database (same server, same index + 100 indexes extra) the indexes DO work and won't be reset.

Op maandag 1 september 2014 08:22:25 UTC+2 schreef Oren Eini:

[Oren Eini (Ayende Rahien)]

No, we haven't pushed the update yet.

[Derek den Haas]

I thought it was this commit: (which was already fixed)
When last collection etags component is initialized we should relay on existing index definitions instead of index stats to prevent from the case that index definition was deleted. - arek

I will test the 3461 version

Op maandag 1 september 2014 08:47:35 UTC+2 schreef Oren Eini:

[Derek den Haas]

Still resetting 2 indexes.

Op maandag 1 september 2014 12:20:47 UTC+2 schreef Derek den Haas:

[Oren Eini (Ayende Rahien)]

On 3461 ? 

Can you go on Skype?

[Derek den Haas]

Yes I can, added you (nickname Oren Eini)

Op maandag 1 september 2014 13:05:47 UTC+2 schreef Oren Eini: