interaction between Virtual server & Authorize section
Hi all, I have two running virtual servers listening different auth port(1645 and 1812), first one for wifi service Authorization and the second for radius check from our NMS by sending simulated auth request. Virtual Server listening on 1645 is connecting to DB by number of sockets, but the other one listening on 1812 doesn't. Just proxy auth request to external AAA and return reply to NMS. My question is: Suppose the DB that one of our virtual server connected is slow, does this virtual server would impact operations of the other virtual server that for NMS? For example, causing other authorized procedures in another virtual server becoming delayed. That is, even virtual server "A" only has few operation to check, while "B" has much more things need to check and unfortunately maybe it could have low performance DB backend, does B might occupy resource in whole Authorize section so that A also become more and more busier? In fact, I got many times of Error msg "child is hung for request in component authorize module" in radius.log. I would not think the loading in our server was really much high enough for causing late response to simulated request from our NMS per *minute*. Our wifi auth request per day is among 2000~3000 in average. So I was curious that is there any mutual influence or interaction between different virtual servers in the same section. If yes, then I think our new DB is highly suspicious. Best regards, Okis.
On Mon, Oct 21, 2013 at 10:53 AM, Okis Chuang <okischuang@outlook.com>wrote:
That is, even virtual server “A” only has few operation to check, while “B” has much more things need to check and unfortunately maybe it could have low performance DB backend, does B might occupy resource in whole Authorize section so that A also become more and more busier?
**
Yes
**
In fact, I got many times of Error msg “child is hung for request in component authorize module” in radius.log. I would not think the loading in our server was really much high enough for causing late response to simulated request from our NMS per **minute**. Our wifi auth request per day is among 2000~3000 in average.
only 2k per day? That's tiny. Something if seriously wrong with your db if it can't handle that much requests. IIRC one way to isolate problematic db is to use small num_sql_socks in sql.conf (default is 5). If the db is slow, it should bomb out quickly with "There are no DB handles to use" errors, but there should be only a small count of FR threads interacting with that db so that processes not using it (e.g. other virtual server) should still work just fine. -- Fajar
participants (2)
-
Fajar A. Nugraha -
Okis Chuang