On 21.12.2019 07:31, Glen wrote: > So just for laughs, I ran an lsmod in both modes, and sorted and diffed them: > > The "clean" guest (which appears to be stable), has these four kernel > modules not present on the upgraded guest: > > iptable_raw > nf_conntrack_ftp > nf_nat_ftp > xt_CT > > The "dup'ped" guest (which seems to be crashable on a large local > rsync) has these modules not present on a clean install: > > auth_rpcgss > br_netfilter > bridge
One of these two is, according to my experience, a fair candidate for your problems. I'm not a networking specialist at all, so I can't give any suggestions on how to convert the upgraded guest to a network config not requiring these modules. (Trying to get rid of br_netfilter alone may be easier, but again I'm not really knowledgeable in this area at all.) Jan > grace > intel_rapl > ipt_MASQUERADE > llc > lockd > nf_conntrack_netlink > nf_nat_masquerade_ipv4 > nfnetlink > nfs_acl > nfsd > overlay > sb_edac > stp > sunrpc > veth > xfrm_algo > xfrm_user > xt_addrtype > xt_nat > > Both guests share these additional sysctl.conf settings: > > kernel.panic = 5 > vm.panic_on_oom = 2 > vm.swappiness = 0 > net.ipv6.conf.all.autoconf = 0 > net.ipv6.conf.default.autoconf = 0 > net.ipv6.conf.eth0.autoconf = 0 > net.ipv4.tcp_fin_timeout = 10 > net.ipv4.tcp_tw_reuse = 0 > > The dup'ped guest has these additional sysctl.conf settings: > > net.ipv4.tcp_tw_recycle = 0 > net.core.netdev_max_backlog=300000 > net.core.somaxconn = 2048 > net.core.rmem_max=67108864 > net.core.wmem_max=67108864 > net.ipv4.ip_local_port_range=15000 65000 > net.ipv4.tcp_sack=0 > net.ipv4.tcp_rmem=4096 87380 67108864 > net.ipv4.tcp_wmem=4096 65536 67108864 > > all of which have, more or less, worked well in the past (when > everything was on 42.3) and may or may not be relevant here. > > I'm sorry, I feel like I'm missing something obvious here, but I can't > see it. I would be grateful for any guidance or insights into this. > Yes, in addition to trying to upgrade my client in place to 15.1, I > could just build a new guest by hand, but that would be even more > time-consuming and seems like it should not be necessary. If I might > quote from the kernel, "Dazed and confused, but trying to continue" is > exactly how I'm feeling here. Why could this guest be hanging? Why > does an NMI bring it back? What should I do next? Anything anyone > would be willing to point me to or suggest would be gratefully > appreciated. > > Glen > -- To unsubscribe, e-mail: [email protected] To contact the owner, e-mail: [email protected]
