6,534
edits
Changes
created (mostly from http://kb.parallels.com/en/117404)
== Changes ==
Since {{kernel link|rhel6|042stab079.6}}:
* Rebase to new RHEL6.4 kernel 2.6.32-358.18.1.el6 (security and bug fixes, see link below)
* Ploop has been enhanced to thoroughly collect a succession of BIOs to improve the performance on small requests (PSBM-19207)
* Performance events are disabled inside the Container by default now. Set 'kernel.perf_event_paranoid' sysctl (default 1) to -1 to enable it back.
* Memory compacting algorithm has been enhanced to work better under high load. (PSBM-20273)
* A kernel panic could occur while accessing a USB storage device. (PSBM-21015)
* Under certain circumstances, signal pending flag could be lost due to a race in the code. (PCLIN-31992)
* The SKB_GSO_TCP flag saving is added to the checkpointing mechanism in order to prevent kernel panic of the destination Node that could happen while the online migration of Containers was in progress. (PSBM-21181)
* Kernel rpm package now provides appropriate base RHEL6 kernel version (like '2.6.32-358.14.1.el6') in the 'kernel' capability. This form of 'kernel' capability is required by the 'irqbalance' rpm package. (PSBM-21258, {{B|2679}})
* Under certain conditions, the file system could be left frozen after the Container backup operation, which caused some processes to stall awaiting for disk I/O. (PCLIN-32011)
* /proc/meminfo output inside a vSwap-enabled Container was extended to include the 'Buffers:' line which is expected by some monitoring tools. Buffers usage is always reported as 0 inside a Container. (PSBM-19448)
* Under certain circumstances, a checkpointing operation of a Container running processes being straced could cause a Hardware Node hard lockup. (PCLIN-32020)
* fsck considers ext4 filesystem broken if it was previously resized online. (PSBM-21709, {{B|2701}})
* Manual unmounting of a ploop device while the inner file system is still being mounted and used by the Container leads to numerous complaints in kernel logs, like: "VFS: Busy inodes after unmount. sb = ffff880108987000, fs type = ext4, sb count = 2, sb->s_root = /". (PSBM-21474)
* The resources accounting has been improved by fixing a number of resources accounting issues.
* OOM berserker mode algorithm (which is a defense from fork bombs) has been tuned to reduce the chances to kill processes from Containers which do not run a fork bomb. (PSBM-21293)
* The Container ID has been added to the SYN flood warning message in order to identify the problem Container on a Hardware Node.(PCLIN-31975)
=== See also ===
* {{RHSA|2013-1173}}
Since {{kernel link|rhel6|042stab079.6}}:
* Rebase to new RHEL6.4 kernel 2.6.32-358.18.1.el6 (security and bug fixes, see link below)
* Ploop has been enhanced to thoroughly collect a succession of BIOs to improve the performance on small requests (PSBM-19207)
* Performance events are disabled inside the Container by default now. Set 'kernel.perf_event_paranoid' sysctl (default 1) to -1 to enable it back.
* Memory compacting algorithm has been enhanced to work better under high load. (PSBM-20273)
* A kernel panic could occur while accessing a USB storage device. (PSBM-21015)
* Under certain circumstances, signal pending flag could be lost due to a race in the code. (PCLIN-31992)
* The SKB_GSO_TCP flag saving is added to the checkpointing mechanism in order to prevent kernel panic of the destination Node that could happen while the online migration of Containers was in progress. (PSBM-21181)
* Kernel rpm package now provides appropriate base RHEL6 kernel version (like '2.6.32-358.14.1.el6') in the 'kernel' capability. This form of 'kernel' capability is required by the 'irqbalance' rpm package. (PSBM-21258, {{B|2679}})
* Under certain conditions, the file system could be left frozen after the Container backup operation, which caused some processes to stall awaiting for disk I/O. (PCLIN-32011)
* /proc/meminfo output inside a vSwap-enabled Container was extended to include the 'Buffers:' line which is expected by some monitoring tools. Buffers usage is always reported as 0 inside a Container. (PSBM-19448)
* Under certain circumstances, a checkpointing operation of a Container running processes being straced could cause a Hardware Node hard lockup. (PCLIN-32020)
* fsck considers ext4 filesystem broken if it was previously resized online. (PSBM-21709, {{B|2701}})
* Manual unmounting of a ploop device while the inner file system is still being mounted and used by the Container leads to numerous complaints in kernel logs, like: "VFS: Busy inodes after unmount. sb = ffff880108987000, fs type = ext4, sb count = 2, sb->s_root = /". (PSBM-21474)
* The resources accounting has been improved by fixing a number of resources accounting issues.
* OOM berserker mode algorithm (which is a defense from fork bombs) has been tuned to reduce the chances to kill processes from Containers which do not run a fork bomb. (PSBM-21293)
* The Container ID has been added to the SYN flood warning message in order to identify the problem Container on a Hardware Node.(PCLIN-31975)
=== See also ===
* {{RHSA|2013-1173}}