hi, I am currently in the process of optimizing my site for speed and memmory usage. Because of this I am experimenting with APC but the size of the {cache} table made me wonder what uses so much space. Part of the Answer was Feeds.
I want to push {cache} to APC but as the feeds caches are only needed for cron runs they should not be in APC. The most obvious way would be to put the feeds caches to a seperate cache bin and exclude this bin from APC.
For this I had to modify the feeds module to use that cache bin.
The attached patch is against the current head and adds a own cache bin for feeds.
I would be interested what you think about this.
thanks
Clemens
Comment | File | Size | Author |
---|---|---|---|
#7 | feeds-cache-table-1449464-7.patch | 3.42 KB | twistor |
#5 | feeds-cache-table-1449464-5.patch | 4 KB | klausi |
feeds_own_cache_bin.diff | 2.95 KB | cbergmann | |
Comments
Comment #1
5n00py CreditAttribution: 5n00py commentedI also have problem with cache table and feeds.
I think this issue is related to next one:
#1360910: Don't cache_set full feed result in http_request.inc (or make optional)
One of the possible solutions is - store only headers in feeds_http_download_ instead of full data.
Comment #2
vinoth.3v CreditAttribution: vinoth.3v commentedAPC always gets 100% fragmentation because of this Feeds.
It will be good if feeds have the own cache bin.
Comment #3
vinoth.3v CreditAttribution: vinoth.3v commentedComment #4
twistor CreditAttribution: twistor commentedThis is 2 parts. One, Feeds should not be caching downloads in the db. There's an issue for this somewhere. Two, Feeds should have a custom cache table since there could be a large number of entries.
Comment #5
klausiThe other issue is #2032333: [Meta] Store remote files in the file system instead of in the database.
We can still start with a separate bin here, which is already an improvement, so that we can take Feeds out of Memcache for example.
Comment #6
twistor CreditAttribution: twistor as a volunteer commentedAwesome, this has been on my mind again lately.
Does it make sense to move the importer digest? It's probably not a big deal either way, but this does get called on hook_node_update.
Comment #7
twistor CreditAttribution: twistor as a volunteer commentedLet's go ahead and make this a dedicated HTTP cache. While we're at it, change the hashing algorithm to make things easier for people who are concerned about that.
Comment #9
twistor CreditAttribution: twistor as a volunteer commentedChanged on commit:
Used sha256 instead of sha512 to be a bit nicer to our indexes.
Fixed comment on feeds_update_7212().