Known Issues¶

This section lists known issues in this release:

• Disk partitioning on controllers does not automatically reset back to the default state if you change the Glance backend. See LP1450100.

• No option to connect to a remote repository via HTTP Proxy without building a local mirror. See LP1460169.

• Default disk allocation may not use entire disk space. To prevent this, before deploying, check that you do not have unallocated disk space by going to Nodes, selecting a node and clicking Configure Disks. See LP1490597.

• A system with CentOS 7.2 and e1000 network interface may become temporarily unresponsive with the system log entries similar to the following ones:

  [ 2045.553869] e1000 0000:00:03.0 enp0s3: Detected Tx Unit Hang
  [ 2045.553869] Tx Queue <0>
  [ 2045.553869] TDH <3>
  [ 2045.553869] TDT <3>
  [ 2045.553869] next_to_use <de>
  [ 2045.553869] next_to_clean <3>
  [ 2045.553869] buffer_info[next_to_clean]
  [ 2045.553869] time_stamp <1001a8c59>
  [ 2045.553869] next_to_watch <16>
  [ 2045.553869] jiffies <1001aa291>
  [ 2045.553869] next_to_watch.status <0>
  [ 2046.720054] ------------[ cut here ]------------
  [ 2046.720084] WARNING: at net/sched/sch_generic.c:297
  dev_watchdog+0x270/0x280()
  [ 2046.720100] NETDEV WATCHDOG: enp0s3 (e1000): transmit queue 0 timed out
  [ 2046.720104] Modules linked in: loop xt_CHECKSUM iptable_mangle nf_log_ipv4
  nf_log_common xt_LOG xt_limit ipt_MASQUERADE nf_nat_masque
  [ 2046.721693] CPU: 0 PID: 0 Comm: swapper/0 Not tainted
  3.10.0-327.3.1.el7.x86_64 #1
  [ 2046.721693] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
  BIOS 1.7.5-20151012_155810-obs-1 04/01/2014
  [ 2046.721693] ffff88007fc03d88 76f0075ca754d7b4 ffff88007fc03d40 ffffffff8163516c
  [ 2046.721693] ffff88007fc03d78 ffffffff8107b200 0000000000000000 ffff88007a352000
  [ 2046.721693] ffff88007878ac80 0000000000000001 0000000000000000 ffff88007fc03de0
  [ 2046.721693] Call Trace:
  [ 2046.721693] <IRQ> [<ffffffff8163516c>] dump_stack+0x19/0x1b
  [ 2046.721693] [<ffffffff8107b200>] warn_slowpath_common+0x70/0xb0
  [ 2046.721693] [<ffffffff8107b29c>] warn_slowpath_fmt+0x5c/0x80
  [ 2046.721693] [<ffffffff8154ca50>] dev_watchdog+0x270/0x280
  [ 2046.721693] [<ffffffff8154c7e0>] ? dev_graft_qdisc+0x80/0x80
  [ 2046.721693] [<ffffffff8108b0a6>] call_timer_fn+0x36/0x110
  [ 2046.721693] [<ffffffff8154c7e0>] ? dev_graft_qdisc+0x80/0x80
  [ 2046.721693] [<ffffffff8108dd97>] run_timer_softirq+0x237/0x340
  [ 2046.721693] [<ffffffff81084b0f>] __do_softirq+0xef/0x280
  [ 2046.721693] [<ffffffff816471dc>] call_softirq+0x1c/0x30
  [ 2046.721693] [<ffffffff81016fc5>] do_softirq+0x65/0xa0
  [ 2046.721693] [<ffffffff81084ea5>] irq_exit+0x115/0x120
  [ 2046.721693] [<ffffffff81647e55>] smp_apic_timer_interrupt+0x45/0x60
  [ 2046.721693] [<ffffffff8164651d>] apic_timer_interrupt+0x6d/0x80
  

  The issue is with both hardware and virtual e1000 network interface. The time during which the system is unresponsive is less than a minute. The cause of the issue is network interface restart. System logs This is a known e1000 driver issue with various distributions: CentOS, Ubuntu, RHEL. See LP1526544.

  Solutions:

  • Use “virtio” driver: To work around the issue with the virtual interface, use the “virtio” driver. This solution may have limitations – for example, bonding interface does not work with “virtio”. Please check if this workaround is compatible with your use case.

  • Disable TSO offloading – run-time only: Disable the setting in run-time with the command:

    ethtool -K <ifname> tso off
    

  • Disable TSO offloading – persistent change: To make the change persistent, create a script that will set this option on every ‘ifup’:

    1. Create an empty file:

       touch /sbin/ifup-local
       

    2. Make the file executable:

       chmod +x /sbin/ifup-local
       

    3. Put the following lines into the file:

       #!/bin/bash
       if [[ -z "${1}" ]]; then
        exit
       fi
       devpath=$(readlink -m /sys/class/net/${1})
       if [[ "${devpath}" == /sys/devices/virtual/* ]]; then
         exit
       fi
       ethtool -K ${1} tso off
       

• Deployment with enabled SR-IOV fails on some hardware with the “sriov_iommu_check” entry in the log. This is a known issue with Linux kernel and some hardware platforms. See LP1556854.

• For performance reasons, integration of S3 API/Keystone is disabled by default. See LP1446704.