Hello,
On nearly all my 4.7 sites, I am having problems with search. The issues are random, but occur on all my sites. After a certain amount of time, results stop being indexed. My search index tells me that all is indexed, though I know, it is not :) When I search for obvious words in newer posts, I get either no results, or too few.
Rebuilding the index works. Untill after a random time, then it stops indexing. Indexing stops on one site a time. This has been going on for months.
Cron runs just fine, so it seems, and I /am/ running multisite sites, with cron runs on all the sites every 16 minutes, so it could be a memory/CPU issue, though I doubt it, since my watchdog tells me cron ran just fine, and my server is far from choked.
Has anyone experienced this too? Has anyone seen any docs on Drupal.org about this specific issue? Does anyone know where to start debugging this weirdness?
Bèr
On Oct 19, 2006, at 1:09 PM, Bèr Kessels wrote:
On nearly all my 4.7 sites, I am having problems with search. The issues are random, but occur on all my sites. After a certain amount of time, results stop being indexed. My search index tells me that all is indexed, though I know, it is not :) When I search for obvious words in newer posts, I get either no results, or too few.
Rebuilding the index works. Untill after a random time, then it stops indexing. Indexing stops on one site a time. This has been going on for months.
I just noticed this on one site I helped migrate to a new server this morning, and figured the table got corrupted. I did not think that it might be a common problem. I'll reply again if I notice it on any other sites.
Laura
Op vrijdag 20 oktober 2006 00:47, schreef Laura Scott:
I just noticed this on one site I helped migrate to a new server this morning, and figured the table got corrupted. I did not think that it might be a common problem. I'll reply again if I notice it on any other sites.
hmm, this seems liek a good point. I ran mysqlcheck -udebian-sys-maint -p --repair --auto-repair --analyze --optimize -A
And it seems that indeed the indexes were corrupted on at least two of the sites that stopped indexing.
Lets hope and see if that was really the root of all problems.
Bèr
On Friday 20 October 2006 06:47, Bèr Kessels wrote:
Op vrijdag 20 oktober 2006 00:47, schreef Laura Scott:
I just noticed this on one site I helped migrate to a new server this morning, and figured the table got corrupted. I did not think that it might be a common problem. I'll reply again if I notice it on any other sites.
hmm, this seems liek a good point. I ran mysqlcheck -udebian-sys-maint -p --repair --auto-repair --analyze --optimize -A
And it seems that indeed the indexes were corrupted on at least two of the sites that stopped indexing.
Lets hope and see if that was really the root of all problems.
Bèr
It would be nice to know why they were corrupt. Was it just solar flares, or is there something in the Drupal cron run (or somewhere else) that is causing it.
On Oct 20, 2006, at 9:15 AM, Jason Flatt wrote:
On Friday 20 October 2006 06:47, Bèr Kessels wrote:
Op vrijdag 20 oktober 2006 00:47, schreef Laura Scott:
I just noticed this on one site I helped migrate to a new server this morning, and figured the table got corrupted. I did not think that it might be a common problem. I'll reply again if I notice it on any other sites.
hmm, this seems liek a good point. I ran mysqlcheck -udebian-sys-maint -p --repair --auto-repair --analyze --optimize -A
And it seems that indeed the indexes were corrupted on at least two of the sites that stopped indexing.
Lets hope and see if that was really the root of all problems.
Bèr
It would be nice to know why they were corrupt. Was it just solar flares, or is there something in the Drupal cron run (or somewhere else) that is causing it.
I just found another site -- my personal blog, actually -- that had searches coming up empty. I re-indexed the site and manually triggere cron a couple of times, and all is working now. I will have to check it now and then, to see if it flakes out on me again.
The two sites I've seen this so far were recently migrated, and I typically don't migrate search indexes, so they were indexed very recently. This is a mystery, but might point to a problem.
Laura
I think we have solved a FAQ problem here! So I am cross posting to the docs ML, hope they can do something with this.
Op zaterdag 21 oktober 2006 00:10, schreef Laura Scott:
The two sites I've seen this so far were recently migrated, and I typically don't migrate search indexes, so they were indexed very recently. This is a mystery, but might point to a problem.
It is not a mistery, and it pointed me to the right place. :)
The progress of search index is maintained separate from the index itself. You can migrate a database, without the index, and Drupal will think indexing is completed, while it is not. Drupal does not acutally check the status and/or health of the index, it noly checks for that progression variable.
That gave me a good lead and I found four different causes that may be the root of all evil: * corrupted tables. repairing seems to have solved this issue on two sites. * cron choking on aggregator importing an invalidated feed (sounds like a core bug to me!). aggregator made cron not finish running. * three tables got 'skewed' during development and migration. Cleaning that out completely, solved the problem. * one site got a 'memory exhausted' during cron. Never seen it appear on the web-front-end, and since cron did not mail/print anything, I never noticed it . "memory exhausted" does not appear on the watchdog.
Op vrijdag 20 oktober 2006 17:15, schreef Jason Flatt:
It would be nice to know why they were corrupt. Was it just solar flares, or is there something in the Drupal cron run (or somewhere else) that is causing it.
Drupal cron, on my server is not optimal configured. I run it every 16 minutes on all sites on that server (100+), which means high peeks at certain random moments. * 16 minutes, because its one-off from 15 and therefore does not hit at the round hour. * 16 minutes, because on most sites in 16 minutes hardl anything happened. make it 1 hour, and you'll see peeks. * all sites at once, because that is easy to maintain in the hosting software. This should avoid peaks, but still, if I put a site online after development, or import a database, or any other off-usual-run thingy, then peaks do occur.
Bèr
-
Bèr Kessels -
Jason Flatt -
Laura Scott