I was wondering about something. For now we do not implement let-say metalinks support integrated inside squid.. Do we have any mechanism that actually do a md5\sha1 etc. check on a specific file that is in cache?
I wrote a small helper long ago that extracts the data of an object from a UFS\AUFS store file. The above also allows iterating over all UFS\AUFS cache_dir files and delete the actual object from cache. The above option can be used to also verify the integrity of a file in the HASH level. The helper can find the object and then verify the object\file data without all the headers and metadata and make sure that it fits MD5\SHA1 hash.. A scenario that I was thinking about was that when StoreID is being used there is a small problem to verify the content of each node in the mirror pool... has the consistent copy of the directory. metalinks can contain info/data on a single and multi files and multi chunks. A admin that wants to use StoreID I believe that will want to know if his setup causing a trouble to the users. A nice Idea I had was to use a "cache problem form" for cases that the user has an issue and then the cache admin can try to investigate the issue. The basic assumption to that is that not all mirrors are updated or maybe in a case of a file corruption from a disk failure or any other issues the admin will want to know about that. Maybe adding a header to the client like "in a case of troubles contact us using xyz" ??(similar to some facebook privacy headers) I want to hear your opinion about it from an angle which is not mine only. pros?cons? I know that for most filtering solutions they do hava a cgi script to report a problems.. Or in other places there is 302 redirection into a specific block page to a form or other option that can help the user and the admins to identify holes or over-filtering scenarios. Eliezer
