Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
We've had some issues with NIC offloads that affected the stability of a lot of different TCP based applications, so I tend to unconditionally disable any NIC offloads on any machine that is transmitting data I care about. I've never seen this negatively impact performance, and almost always improve things for TCP based applications like DRBD. It could also be why the newer kernels seem affected, as they've likely implemented or enabled more of the native offloading in different NICs. Every one of our DRBD clusters have all their NIC offloads disabled and we've had no issues with them at all. You could try the same and see how it works for you since it seems to be NIC related for you. I know there have been lots of issues recently, not just DRBD, with NIC offloading and applications breaking. Cheers, Joel On 04/10/2015 03:27 PM, Jean-Francois Malouin wrote: > Hi, > > Allow me to add my grain of salt to this: sorry if it's too long. > We have a system exibiting the same behaviour and it is certainly not > DRBD-related as it doesn't use it. > > It runs Xen 4.4.2 though with kernel 3.19.3. > And we have seen the same behaviour with 3.14.36 and 3.18.10. > Xen and kernel are in-house compiled. > > Symptoms: before either simply crashing or starting being unresponsive > network-wize. A few instances: > > Apr 7 15:33:10 kernel: [459694.670896] ------------[ cut here ]------------ > Apr 7 15:33:10 kernel: [459694.670921] WARNING: CPU: 0 PID: 0 at net/sched/sch_generic.c:303 dev_watchdog+0x165/0x20d() > Apr 7 15:33:10 kernel: [459694.670924] NETDEV WATCHDOG: eth0 (igb): transmit queue 6 timed out > Apr 7 15:33:10 agrippa kernel: [459694.670927] Modules linked in: st nfsv3 nfsv4 xt_physdev br_netfilter autofs4 parport_pc af_packet ppdev lp xen_acpi_processor parport xen_netback xen_blkback xen_gntalloc fuse rpcsec_gss_krb5 nfsd auth_rpcgss oid_registry nfs_acl nfs lockd grace fscache sunrpc bridge stp llc ipv6 ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat nf_nat_ipv4 nf_nat ipt_REJECT nf_reject_ipv4 nf_conntr ack_ipv4 nf_defrag_ipv4 xt_state nf_conntrack xt_tcpudp iptable_filter iptable_mangle ip_tables x_tables ipmi_devintf iTCO_wdt iTCO_vendor_support coretemp pcspkr i2c_i801 evbug evdev joydev lpc_ich i7core_edac ioatdma edac_core i5500_temp rtc_cmos ipmi_si tpm_tis ipmi_msghandler tpm button processor usbkbd usbmouse usbhid sg igb i2c_algo_bit ehci_pci uhci_hcd i2c_core ehci_hcd usbcore ixgbe dca e1000e hwmon ptp pps_co re mdio megaraid_sas > Apr 7 15:33:10 agrippa kernel: [459694.671006] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.19.3-i686-64-smp #1 > Apr 7 15:33:10 agrippa kernel: [459694.671009] Hardware name: Supermicro X8DTU-6+/X8DTU-6+, BIOS 2.1b 11/15/2011 > > > For 3.14.33: > > Mar 15 05:45:43 agrippa kernel: [1612129.231013] ------------[ cut here ]------------ > Mar 15 05:45:43 agrippa kernel: [1612129.231024] WARNING: CPU: 14 PID: 0 at net/sched/sch_generic.c:264 dev_watchdog+0x161/0x216() > Mar 15 05:45:43 agrippa kernel: [1612129.231027] NETDEV WATCHDOG: eth0 (igb): transmit queue 2 timed out > Mar 15 05:45:43 agrippa kernel: [1612129.231029] Modules linked in: x25 appletalk ipx p8023 p8022 psnap rose netrom ax25 ipt_MASQUERADE xt_state iptable_mangle xt_physdev st ipt_REJECT bridge stp llc nfsv3 nfsv4 autofs4 parport_pc ppdev lp parport af_packet xen_acpi_processor xen_netback xen_blkback xen_gntalloc fuse rpcsec_gss_krb5 nfsd auth_rpcgss oid_registry nfs_acl nfs lockd fscache sunrpc ipv6 iptable_filter xt_n at xt_tcpudp iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack ip_tables x_tables ipmi_devintf iTCO_wdt iTCO_vendor_support coretemp microcode pcspkr evbug evdev joydev i2c_i801 lpc_ic h i7core_edac ioatdma edac_core rtc_cmos tpm_tis tpm ipmi_si ipmi_msghandler button processor usbkbd usbmouse usbhid sg ehci_pci igb uhci_hcd ehci_hcd i2c_algo_bit i2c_core usbcore ixgbe e1000e dca hwmon ptp pps _core megaraid_sas mdio > Mar 15 05:45:43 agrippa kernel: [1612129.231118] CPU: 14 PID: 0 Comm: swapper/14 Not tainted 3.14.33-i686-64-smp #1 > Mar 15 05:45:43 agrippa kernel: [1612129.231120] Hardware name: Supermicro X8DTU-6+/X8DTU-6+, BIOS 2.1b 11/15/2011 > > > This happens around one or twice a week. Absolutely very annoying. We > are still investigating weather it is related specifically to the > onboard Intel Gigabit adapter. Or not. > > Board Manufacturer: Supermicro, Product Name: X8DTU-6+ > > ~# modinfo igb > filename: > /lib/modules/3.19.3-i686-64-smp/kernel/drivers/net/ethernet/intel/igb/igb.ko > version: 5.2.15-k > license: GPL > description: Intel(R) Gigabit Ethernet Network Driver > author: Intel Corporation, <e1000-devel at lists.sourceforge.net> > [..%<..snip..%<..] > > HTH, > jf > > > * Igor Novgorodov <igor at novg.net> [20150410 15:32]: >> I'd second previous poster - try another kernel, long-term-supported >> 3.18 preferably. >> DRBD by itself is rock solid, your problems with timeouts lie elsewhere. >> I've been running DRBD server pairs on custom built 3.14 LTS kernel >> with a year+ uptimes, no problems at all. >> >> And don't use Intel's out-of-kernel drivers if you don't clearly see >> you need it. In-kernel ones very good >> and will surely work more stable. >> >> So, concluding: >> 1. LTS kernel >> 2. In-kernel DRBD >> 3. In-kernel drivers >> 4. Maybe try to stress-test the system without Xen, only DRBD stuff >> (run fio tool in random read-write for a couple of days on DRBD >> device). >> 5. Don't use unstable distributions, like Jessie, it's better to >> backport the needed stuff to Wheezy if it's not already in >> wheezy-backports repo. >> >> On 09/04/15 23:07, Alan Evetts wrote: >>> Hi there, >>> >>> I am reaching out because we have been trying to find stability in our move to DRBD as it is amazing in concept, but have struggled for 6 months of time. I am going to just lay out everything we are doing, as the problem starts and stops when we introduce/remove DRBD from the picture. Obviously, these setups get complicated so hopefully this isn’t too much information here. >>> >>> What we are trying to do is have a pair of Dell R610 machines, each running DRBD and xen with about 8 DRBD partitions, each master running half of the Xen virtual machines. >>> >>> Seems, between 1 and 20 days we always receive a kernel panic on 1 machine, which will often drag down the second machine. Details of the most recent panic are below. >>> >>> In order to rule out problems we have: >>> - Replace both Dell R610 (have 4 now total, all the same problem) >>> - Upgraded to Debian Jessie from Debian Wheezy >>> - Running xen-hypervisor-4.4-amd64, drbd debian version 8.9.2~rc1-2, kernel 3.16.0-4 >>> - Switched from the on-board broadcom NICs to Intel E1G44HTBLK 4 port PCI-e NIC >>> - Upgraded to igb kernel module 5.2.17 and rebuilt it into the initrd as well >>> >>> >>> The 2 servers both have lots of resources (64 gigs of ram, quad xeon 2.4, 6 * 1 TB drives in a raid 10). There is a cross over cable on ETH3 for DRBD, each drbd instance runs on its own port on ETH3. The Xen config runs on a bridge. >>> >>> The problem has more or less been the same as we’ve moved through all of the hardware and software versions over the past 6 months. It rotates between the servers. >>> >>> I am hoping someone can spot a problem in our config, or guide us on what to try from here. All 4 dell machines have been patched and had the diagnostics ran on them without issue. >>> >>> The problem. One of the machines will have a transit queue time-out on an interface (oddly, not necessarily the drbd interface - but usually). From there, a panic, and the NIC will start going up and down. This then starts to drive the load up, the machines soon become unresponsive over shell. Connected over the dRAC remote access port, sooner or later we see errors about the drives not responding, I think this is from the load but I do not know for sure. From this point the machine will sometimes drag down its paired DRBD machine, and sometimes not. The one with the crash needs a hard reboot at this point. >>> >>> We love DRBD, its simplicity and functionality but it introduces these often crashes which are not worth it. Hoping someone can spot an error we are doing here, or have ideas on what to try. >>> >>> Thanks in advance for any help.. and FYI this crashed used to happen in the broadcom queue, now its the intel queue, and only when we have drbd enabled. >>> >>> >>> >>> Apr 9 03:39:17 v2 kernel: [141714.850432] ------------[ cut here ]------------ >>> Apr 9 03:39:17 v2 kernel: [141714.850521] WARNING: CPU: 0 PID: 0 at /build/linux-y7bjb0/linux-3.16.7-ckt4/net/sched/sch_generic.c:264 dev_watchdog+0x236/0x240() >>> Apr 9 03:39:17 v2 kernel: [141714.850527] NETDEV WATCHDOG: eth1 (igb): transmit queue 0 timed out >>> Apr 9 03:39:17 v2 kernel: [141714.850531] Modules linked in: xt_tcpudp xt_physdev iptable_filter ip_tables x_tables xen_netback xen_blkback nfnetlink_queue nfnetlink_log nfnetlink bluetooth 6lowpan_iphc rfkill xen_gntdev xen_evt >>> chn xenfs xen_privcmd nfsd auth_rpcgss oid_registry nfs_acl nfs lockd fscache sunrpc bridge stp llc ttm drm_kms_helper joydev drm i2c_algo_bit i2c_core pcspkr wmi iTCO_wdt iTCO_vendor_support psmouse dcdbas serio_raw evdev tpm_ti >>> s tpm lpc_ich mfd_core acpi_power_meter button coretemp i7core_edac edac_core shpchp processor thermal_sys loop ipmi_watchdog ipmi_si ipmi_poweroff ipmi_devintf ipmi_msghandler drbd lru_cache libcrc32c autofs4 ext4 crc16 mbcache >>> jbd2 dm_mod sg sd_mod crc_t10dif crct10dif_generic sr_mod cdrom ses crct10dif_common enclosure ata_generic hid_generic usbhid hid crc32c_intel ata_piix ehci_pci uhci_hcd libata igb(O) megaraid_sas ehci_hcd scsi_mod usbcore dca pt >>> p usb_common pps_core >>> Apr 9 03:39:17 v2 kernel: [141714.850609] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G O 3.16.0-4-amd64 #1 Debian 3.16.7-ckt4-3 >>> Apr 9 03:39:17 v2 kernel: [141714.850613] Hardware name: Dell Inc. PowerEdge R610/0XDN97, BIOS 6.4.0 07/23/2013 >>> Apr 9 03:39:17 v2 kernel: [141714.850617] 0000000000000009 ffffffff815096a7 ffff880079e03e28 ffffffff810676f7 >>> Apr 9 03:39:17 v2 kernel: [141714.850622] 0000000000000000 ffff880079e03e78 0000000000000010 0000000000000000 >>> Apr 9 03:39:17 v2 kernel: [141714.850626] ffff8800445c8000 ffffffff8106775c ffffffff81777270 ffffffff00000030 >>> Apr 9 03:39:17 v2 kernel: [141714.850631] Call Trace: >>> Apr 9 03:39:17 v2 kernel: [141714.850635] <IRQ> [<ffffffff815096a7>] ? dump_stack+0x41/0x51 >>> Apr 9 03:39:17 v2 kernel: [141714.850652] [<ffffffff810676f7>] ? warn_slowpath_common+0x77/0x90 >>> Apr 9 03:39:17 v2 kernel: [141714.850660] [<ffffffff8106775c>] ? warn_slowpath_fmt+0x4c/0x50 >>> Apr 9 03:39:17 v2 kernel: [141714.850669] [<ffffffff81074647>] ? mod_timer+0x127/0x1e0 >>> Apr 9 03:39:17 v2 kernel: [141714.850676] [<ffffffff8143ce76>] ? dev_watchdog+0x236/0x240 >>> Apr 9 03:39:17 v2 kernel: [141714.850681] [<ffffffff8143cc40>] ? dev_graft_qdisc+0x70/0x70 >>> Apr 9 03:39:17 v2 kernel: [141714.850686] [<ffffffff810729b1>] ? call_timer_fn+0x31/0x100 >>> Apr 9 03:39:17 v2 kernel: [141714.850691] [<ffffffff8143cc40>] ? dev_graft_qdisc+0x70/0x70 >>> Apr 9 03:39:17 v2 kernel: [141714.850698] [<ffffffff81073fe9>] ? run_timer_softirq+0x209/0x2f0 >>> Apr 9 03:39:17 v2 kernel: [141714.850704] [<ffffffff8106c591>] ? __do_softirq+0xf1/0x290 >>> Apr 9 03:39:17 v2 kernel: [141714.850709] [<ffffffff8106c965>] ? irq_exit+0x95/0xa0 >>> Apr 9 03:39:17 v2 kernel: [141714.850718] [<ffffffff813579c5>] ? xen_evtchn_do_upcall+0x35/0x50 >>> Apr 9 03:39:17 v2 kernel: [141714.850725] [<ffffffff8151141e>] ? xen_do_hypervisor_callback+0x1e/0x30 >>> Apr 9 03:39:17 v2 kernel: [141714.850728] <EOI> [<ffffffff810013aa>] ? xen_hypercall_sched_op+0xa/0x20 >>> Apr 9 03:39:17 v2 kernel: [141714.850737] [<ffffffff810013aa>] ? xen_hypercall_sched_op+0xa/0x20 >>> Apr 9 03:39:17 v2 kernel: [141714.850746] [<ffffffff81009e0c>] ? xen_safe_halt+0xc/0x20 >>> Apr 9 03:39:17 v2 kernel: [141714.850756] [<ffffffff8101c959>] ? default_idle+0x19/0xb0 >>> Apr 9 03:39:17 v2 kernel: [141714.850764] [<ffffffff810a7dc0>] ? cpu_startup_entry+0x340/0x400 >>> Apr 9 03:39:17 v2 kernel: [141714.850770] [<ffffffff81902071>] ? start_kernel+0x492/0x49d >>> Apr 9 03:39:17 v2 kernel: [141714.850775] [<ffffffff81901a04>] ? set_init_arg+0x4e/0x4e >>> Apr 9 03:39:17 v2 kernel: [141714.850781] [<ffffffff81903f64>] ? xen_start_kernel+0x569/0x573 >>> Apr 9 03:39:17 v2 kernel: [141714.850785] ---[ end trace ee11063cf033829a ]--- >>> Apr 9 03:39:17 v2 kernel: [141714.871945] br1: port 1(eth1) entered disabled state >>> Apr 9 03:39:20 v2 kernel: [141718.210743] igb 0000:05:00.1 eth1: igb: eth1 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None >>> Apr 9 03:39:20 v2 kernel: [141718.210913] br1: port 1(eth1) entered forwarding state >>> Apr 9 03:39:20 v2 kernel: [141718.210923] br1: port 1(eth1) entered forwarding state >>> Apr 9 03:39:26 v2 kernel: [141723.863194] br1: port 1(eth1) entered disabled state >>> Apr 9 03:39:30 v2 kernel: [141727.650897] igb 0000:05:00.1 eth1: igb: eth1 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None >>> Apr 9 03:39:30 v2 kernel: [141727.651040] br1: port 1(eth1) entered forwarding state >>> Apr 9 03:39:30 v2 kernel: [141727.651053] br1: port 1(eth1) entered forwarding state >>> Apr 9 03:39:31 v2 kernel: [141728.890509] ata1: lost interrupt (Status 0x50) >>> Apr 9 03:39:31 v2 kernel: [141728.890560] sr 1:0:0:0: CDB: >>> Apr 9 03:39:31 v2 kernel: [141728.890563] Get event status notification: 4a 01 00 00 10 00 00 00 08 00 >>> Apr 9 03:39:31 v2 kernel: [141728.890630] ata1: hard resetting link >>> Apr 9 03:39:31 v2 kernel: [141729.366592] ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300) >>> Apr 9 03:39:32 v2 kernel: [141729.406749] ata1.00: configured for UDMA/100 >>> Apr 9 03:39:32 v2 kernel: [141729.408192] ata1: EH complete >>> Apr 9 03:39:35 v2 kernel: [141732.711653] br1: port 1(eth1) entered disabled state >>> Apr 9 03:39:37 v2 kernel: [141734.678485] drbd s3: peer( Primary -> Unknown ) conn( Connected -> NetworkFailure ) pdsk( UpToDate -> DUnknown ) >>> Apr 9 03:39:37 v2 kernel: [141734.678846] drbd s3: asender terminated >>> Apr 9 03:39:37 v2 kernel: [141734.678852] drbd s3: Terminating drbd_a_s3 >>> Apr 9 03:39:37 v2 kernel: [141734.678956] drbd s3: Connection closed >>> Apr 9 03:39:37 v2 kernel: [141734.678972] drbd s3: conn( NetworkFailure -> Unconnected ) >>> Apr 9 03:39:37 v2 kernel: [141734.678974] drbd s3: receiver terminated >>> Apr 9 03:39:37 v2 kernel: [141734.678976] drbd s3: Restarting receiver thread >>> Apr 9 03:39:37 v2 kernel: [141734.678977] drbd s3: receiver (re)started >>> Apr 9 03:39:37 v2 kernel: [141734.678987] drbd s3: conn( Unconnected -> WFConnection ) >>> Apr 9 03:39:38 v2 kernel: [141735.718898] igb 0000:05:00.1 eth1: igb: eth1 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None >>> Apr 9 03:39:38 v2 kernel: [141735.719086] br1: port 1(eth1) entered forwarding state >>> Apr 9 03:39:38 v2 kernel: [141735.719095] br1: port 1(eth1) entered forwarding state >>> Apr 9 03:39:39 v2 kernel: [141737.154575] drbd s4: peer( Secondary -> Unknown ) conn( Connected -> NetworkFailure ) pdsk( UpToDate -> DUnknown ) >>> Apr 9 03:39:39 v2 kernel: [141737.154671] block drbd1: new current UUID 461FF401E0489AAB:9279A3BA4A3A710B:0E977CC4BB5727A9:0E967CC4BB5727A9 >>> Apr 9 03:39:39 v2 kernel: [141737.154921] drbd s4: asender terminated >>> Apr 9 03:39:39 v2 kernel: [141737.154928] drbd s4: Terminating drbd_a_s4 >>> Apr 9 03:39:39 v2 kernel: [141737.155289] drbd s4: Connection closed >>> Apr 9 03:39:39 v2 kernel: [141737.155579] drbd s4: conn( NetworkFailure -> Unconnected ) >>> Apr 9 03:39:39 v2 kernel: [141737.155583] drbd s4: receiver terminated >>> Apr 9 03:39:39 v2 kernel: [141737.155585] drbd s4: Restarting receiver thread >>> Apr 9 03:39:39 v2 kernel: [141737.155586] drbd s4: receiver (re)started >>> Apr 9 03:39:39 v2 kernel: [141737.155601] drbd s4: conn( Unconnected -> WFConnection ) >>> Apr 9 03:39:41 v2 kernel: [141738.458578] drbd n5: peer( Secondary -> Unknown ) conn( Connected -> NetworkFailure ) pdsk( UpToDate -> DUnknown ) >>> Apr 9 03:39:41 v2 kernel: [141738.458671] block drbd8: new current UUID 808265F24E5A3F21:B63FFF468380B383:240D9C7D536ACB97:240C9C7D536ACB97 >>> Apr 9 03:39:41 v2 kernel: [141738.458885] drbd n5: asender terminated >>> Apr 9 03:39:41 v2 kernel: [141738.458893] drbd n5: Terminating drbd_a_n5 >>> Apr 9 03:39:41 v2 kernel: [141738.459160] drbd n5: Connection closed >>> Apr 9 03:39:41 v2 kernel: [141738.459316] drbd n5: conn( NetworkFailure -> Unconnected ) >>> Apr 9 03:39:41 v2 kernel: [141738.459319] drbd n5: receiver terminated >>> Apr 9 03:39:41 v2 kernel: [141738.459321] drbd n5: Restarting receiver thread >>> Apr 9 03:39:41 v2 kernel: [141738.459322] drbd n5: receiver (re)started >>> Apr 9 03:39:41 v2 kernel: [141738.459336] drbd n5: conn( Unconnected -> WFConnection ) >>> Apr 9 03:39:44 v2 kernel: [141742.202552] drbd r1: peer( Primary -> Unknown ) conn( Connected -> NetworkFailure ) pdsk( UpToDate -> DUnknown ) >>> Apr 9 03:39:44 v2 kernel: [141742.202913] drbd r1: asender terminated >>> Apr 9 03:39:44 v2 kernel: [141742.202920] drbd r1: Terminating drbd_a_r1 >>> Apr 9 03:39:44 v2 kernel: [141742.203023] drbd r1: Connection closed >>> Apr 9 03:39:44 v2 kernel: [141742.203039] drbd r1: conn( NetworkFailure -> Unconnected ) >>> Apr 9 03:39:44 v2 kernel: [141742.203041] drbd r1: receiver terminated >>> Apr 9 03:39:44 v2 kernel: [141742.203043] drbd r1: Restarting receiver thread >>> Apr 9 03:39:44 v2 kernel: [141742.203044] drbd r1: receiver (re)started >>> Apr 9 03:39:44 v2 kernel: [141742.203054] drbd r1: conn( Unconnected -> WFConnection ) >>> >>> >>> >>> Etc. >>> >>> >>> >>> >>> _______________________________________________ >>> drbd-user mailing list >>> drbd-user at lists.linbit.com >>> http://lists.linbit.com/mailman/listinfo/drbd-user >> >> _______________________________________________ >> drbd-user mailing list >> drbd-user at lists.linbit.com >> http://lists.linbit.com/mailman/listinfo/drbd-user > _______________________________________________ > drbd-user mailing list > drbd-user at lists.linbit.com > http://lists.linbit.com/mailman/listinfo/drbd-user -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20150410/2f138ce0/attachment.htm>