elsen@kulesat.uucp (07/08/88)
- Does anyone know in what circumstances ones OVERDRAFT diskquota can be used ? I noticed the sometimes LSE for instance can allocate 'overdraft' blocks but I was wondering whether there exists some general rule. - In our DECnet area we configured 2 area routers.I noticed that when the second area router came into operation the NETACP processes on both nodes consumed much more permanent CPU. A '$ sho proc/cont' also revealed a dramatic increase in permanet BUF IO count. Has anyone noticed this before or has an explanation for it ? Marc Elsen (System Manager/Software Engineer) Kathol. Univ. of Louvain Dep. E.S.A.T. Kard. Mercierlaan 94 3030 HEVERLEE Belgium tel. 32(0)16220931(ext. 1080) EMAIL : ...!kulcs!kulesat!elsen (UUCP) elsen%kulesat@uunet.uu.net (via UUNET) elsen%kulesat.uucp@blekul60 (BITNET) psi%020622166012::elsen (PSI MAIL)
LEICHTER@Venus.YCC.Yale.EDU ("Jerry Leichter ", LEICHTER-JERRY@CS.YALE.EDU) (07/12/88)
- Does anyone know in what circumstances ones OVERDRAFT diskquota can be used? I noticed the sometimes LSE for instance can allocate 'overdraft' blocks but I was wondering whether there exists some general rule. The basic rule is simple, if unstated anywhere I could find it: When an attempt to extend an already-open file exceeds the permanent quota, but is within the overdraft, the extension is made anyway, and an alternative status is returned to the caller. I haven't been able to figure out from the documentation what the status code, or even the severity, of the status returned (a) from the ACP EXTEND request or (b) from RMS if it receives that value back form the ACP. Best guess is OVRDSKQUOTA, "Disk usage exceeds disk quota", which is an alternate success return. (This is a value returned by the ACP or SYSTEM.) An ACP EXTEND request can be made as part of a MODIFY request (for an existing file), or as part of a CREATE. As far as I can determine, MODIFY with EXTEND will return this value when it crosses the boundary between permanent and overdraft allocation, and CREATE will refuse to do it, so you can't create any new files in this domain. Once you are into the OVERDRAFT domain, MODIFY/EXTEND works normally, but CREATE/EXTEND fails. Vague as the internal details are, what the USER sees is harder to pin down, since it depends ENTIRELY on what the program he is running chooses to do with an alternate return code. Many programs seem unable to deal with this situa- tion and simply roll over and die. TPU (on which LSE is built) handles it gracefully: It tells you what is going on, but then continues as long as it can. - In our DECnet area we configured 2 area routers.I noticed that when the second area router came into operation the NETACP processes on both nodes consumed much more permanent CPU. A '$ sho proc/cont' also revealed a dramatic increase in permanet BUF IO count. Has anyone noticed this before or has an explanation for it ? You haven't really described your full setup. Are these routers for the SAME area or for DIFFERENT areas? Did bringing up the second area router establish connections to yet more areas "on its other side"? In any case, I'd certainly expect SOME additional work when you have two area routers: With only one, there is no inter-area-router traffic to handle; with two, there is. How much to expect is hard to say. Before you spend too much time chasing this, you might consider whether the "dramatic increase" in CPU and I/O that you saw is having any significant effect on your system. If it is, it might be worth tracking down; if it's not, why waste your time? -- Jerry