Hi all,
we have some problems using server-server virtual volumes, maybe some
of You have seen this before:
We have 2 ADSM servers (RS6K-S70, AIX 4.3.2, ADSM 3.1.2.42) which are
backing up their storage pools into s-s storage pools of the other
server, some ten thousand objects and 100 Gigs of data each. s-s
Volume size is 1G. The servers are connected to each other via direct
full-duplex 100tx ethernet.
When reclamation and/or expiration is running on both servers (any
combination), after some time both servers are stopping receiving
data. No error msg, but no further data accepted, and both recl/exp
processes are not processing. Conclusion: all processing stopped but
the dsmadmc is usable, except of (sometimes) q mount and q stgp
requests. It seems to me that the servers are going into deadlock when
s-s volumes are deleted on the target server, where some exp/recl
processing is deleting some other objects.
When I cancel some cross-server sessions processing continues,
sometimes until the next lock occurs.
There is NO error msg in q actl and sh deadlock shows nothing.
We noticed this on lvl 3.1.2.20 and with 3.1.2.42 it still remains...
Thanx
Roger Hohmann
WestLB
Herzogstr 15
D-40217 D|sseldorf
mailto: roger_hohmann AT westlb DOT de
phone: ++49 211 826 8155
|