[opensuse] Problem with system suddenly swapping with large amount of ram (seems a memory leak?)
Since the update to Leap 15, the longer my system runs, the more memory it consumes. I tried to track this but I do not understand. Following is the output of the terminal of meminfo and I join also a link with the output (snapshot) of top. https://paste.opensuse.org/16233536 Note that the system has 8 gb of RAM and should never swap. Currently it does swap until it blocks. If you restart the swap returns to normal..for some time. Then, (seems related with the time that it runs therefore I thought of a memory leak) it rebuilds up the swap demand. sudo egrep --color 'Mem|Cache|Swap' /proc/meminfo [sudo] password di root: MemTotal: 7953176 kB MemFree: 137320 kB MemAvailable: 417340 kB Cached: 524404 kB SwapCached: 201472 kB SwapTotal: 8388604 kB SwapFree: 5495036 kB entropia@hk9l:~> sudo cat /proc/meminfo MemTotal: 7953176 kB MemFree: 132844 kB MemAvailable: 433172 kB Buffers: 29772 kB Cached: 582808 kB SwapCached: 202804 kB Active: 6208760 kB Inactive: 1211252 kB Active(anon): 6011452 kB Inactive(anon): 914676 kB Active(file): 197308 kB Inactive(file): 296576 kB Unevictable: 128 kB Mlocked: 128 kB SwapTotal: 8388604 kB SwapFree: 5495804 kB Dirty: 300 kB Writeback: 0 kB AnonPages: 6748820 kB Mapped: 294356 kB Shmem: 190476 kB Slab: 153200 kB SReclaimable: 54828 kB SUnreclaim: 98372 kB KernelStack: 22580 kB PageTables: 165200 kB NFS_Unstable: 0 kB Bounce: 0 kB WritebackTmp: 0 kB CommitLimit: 12365192 kB Committed_AS: 15205204 kB VmallocTotal: 34359738367 kB VmallocUsed: 0 kB VmallocChunk: 0 kB HardwareCorrupted: 0 kB AnonHugePages: 460800 kB ShmemHugePages: 0 kB ShmemPmdMapped: 0 kB HugePages_Total: 0 HugePages_Free: 0 HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize: 2048 kB DirectMap4k: 315452 kB DirectMap2M: 7864320 kB _________________________________________________________________ ________________________________________________________ Ihre E-Mail-Postf�cher sicher & zentral an einem Ort. Jetzt wechseln und alte E-Mail-Adresse mitnehmen! https://www.eclipso.de -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 8 June 2018 at 15:16, stakanov <stakanov@eclipso.eu> wrote:
Since the update to Leap 15, the longer my system runs, the more memory it consumes. I tried to track this but I do not understand. Following is the output of the terminal of meminfo and I join also a link with the output (snapshot) of top.
https://paste.opensuse.org/16233536
Note that the system has 8 gb of RAM and should never swap. Currently it does swap until it blocks. If you restart the swap returns to normal..for some time. Then, (seems related with the time that it runs therefore I thought of a memory leak) it rebuilds up the swap demand.
3175 mercurio 20 0 10,122g 3,699g 4492 S 0,000 48,76 2:38.62 kontact 48.76% of your RAM being used by kontact I'm pretty sure that's your leaking process -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
Sorry Richard that went out as PM against my intention. So again to the list as it would have been correct. My apologies. In data venerdì 8 giugno 2018 15:22:35 CEST, hai scritto:
On 8 June 2018 at 15:16, stakanov <stakanov@eclipso.eu> wrote:
Since the update to Leap 15, the longer my system runs, the more memory it consumes. I tried to track this but I do not understand. Following is the output of the terminal of meminfo and I join also a link with the output (snapshot) of top.
https://paste.opensuse.org/16233536
Note that the system has 8 gb of RAM and should never swap. Currently it does swap until it blocks. If you restart the swap returns to normal..for some time. Then, (seems related with the time that it runs therefore I thought of a memory leak) it rebuilds up the swap demand.
3175 mercurio 20 0 10,122g 3,699g 4492 S 0,000 48,76 2:38.62 kontact
48.76% of your RAM being used by kontact
I'm pretty sure that's your leaking process
I was not sure by I think you confirm my doubts. Does this warrant a bug report? Against leap or against KDE directly. And most importantly, how can one know what component of kontact is leaking? Sorry that I ask all this but I want to avoid useless reports. Thanks in advance. BTW, I tried to close kontact with "exit" but the swap stays once it accumulated. Do I have to issue a command to "unswap" the memory? _________________________________________________________________ ________________________________________________________ Ihre E-Mail-Postfächer sicher & zentral an einem Ort. Jetzt wechseln und alte E-Mail-Adresse mitnehmen! https://www.eclipso.de -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 06/08/2018 08:22 AM, Richard Brown wrote:
3175 mercurio 20 0 10,122g 3,699g 4492 S 0,000 48,76 2:38.62 kontact
48.76% of your RAM being used by kontact
I'm pretty sure that's your leaking process
Yes No. 265 does stick out like a sore-thumb. -- David C. Rankin, J.D.,P.E. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
stakanov wrote:
Since the update to Leap 15, the longer my system runs, the more memory it consumes. I tried to track this but I do not understand. Following is the output of the terminal of meminfo and I join also a link with the output (snapshot) of top.
Just for comparison - my leap15 KDE desktop - not overly active, browser, gimp, skanlite - runs 185 processes. 4G of memory of which 3.5G used. Yours is running more than twice the number processes.
Note that the system has 8 gb of RAM and should never swap.
With that workload, yeah, it might very well. Seems to me you have a lot of akonadi stuff running? -- Per Jessen, Zürich (25.4°C) http://www.dns24.ch/ - your free DNS host, made in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
In data venerdì 8 giugno 2018 15:51:54 CEST, Per Jessen ha scritto:
stakanov wrote:
Since the update to Leap 15, the longer my system runs, the more memory it consumes. I tried to track this but I do not understand. Following is the output of the terminal of meminfo and I join also a link with the output (snapshot) of top.
Just for comparison - my leap15 KDE desktop - not overly active, browser, gimp, skanlite - runs 185 processes. 4G of memory of which 3.5G used. Yours is running more than twice the number processes.
Note that the system has 8 gb of RAM and should never swap.
With that workload, yeah, it might very well.
Seems to me you have a lot of akonadi stuff running? akonadi went nuts over the update. I have only two users with kontact / akonadi running that is mercurio and entropia, where entropia does imap and mailinglists and mercurio only email. No fancy stuff. Somewhere in the update something went not really as it should have? But it is definitively some leak as it grows endlessly if let alone in idle.
_________________________________________________________________ ________________________________________________________ Ihre E-Mail-Postfächer sicher & zentral an einem Ort. Jetzt wechseln und alte E-Mail-Adresse mitnehmen! https://www.eclipso.de -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On June 8, 2018 6:54:55 AM PDT, stakanov <stakanov@eclipso.eu> wrote:
In data venerdì 8 giugno 2018 15:51:54 CEST, Per Jessen ha scritto:
stakanov wrote:
Since the update to Leap 15, the longer my system runs, the more memory it consumes. I tried to track this but I do not understand. Following is the output of the terminal of meminfo and I join also a link with the output (snapshot) of top.
Just for comparison - my leap15 KDE desktop - not overly active, browser, gimp, skanlite - runs 185 processes. 4G of memory of which 3.5G used. Yours is running more than twice the number processes.
Note that the system has 8 gb of RAM and should never swap.
With that workload, yeah, it might very well.
Seems to me you have a lot of akonadi stuff running? akonadi went nuts over the update. I have only two users with kontact /
akonadi running that is mercurio and entropia, where entropia does imap and mailinglists and mercurio only email. No fancy stuff. Somewhere in the update something went not really as it should have? But it is definitively some leak as it grows endlessly if let alone in idle.
Maybe look into akonadictl restart or something. Also Kontact is far far from ready for prime time in a production environment. -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On June 8, 2018 6:54:55 AM PDT, stakanov <stakanov@eclipso.eu> wrote:
In data venerdì 8 giugno 2018 15:51:54 CEST, Per Jessen ha scritto:
stakanov wrote:
Since the update to Leap 15, the longer my system runs, the more memory it consumes. I tried to track this but I do not understand. Following is the output of the terminal of meminfo and I join also
a
link with the output (snapshot) of top.
Just for comparison - my leap15 KDE desktop - not overly active, browser, gimp, skanlite - runs 185 processes. 4G of memory of which 3.5G used. Yours is running more than twice the number processes.
Note that the system has 8 gb of RAM and should never swap.
With that workload, yeah, it might very well.
Seems to me you have a lot of akonadi stuff running?
akonadi went nuts over the update. I have only two users with kontact /
akonadi running that is mercurio and entropia, where entropia does imap and mailinglists and mercurio only email. No fancy stuff. Somewhere in the update something went not really as it should have? But it is definitively some leak as it grows endlessly if let alone in idle.
Maybe look into akonadictl restart or something. Also Kontact is far far from ready for prime time in a production environment. I tried with akonadictl restart, or with akonadictl stop I have narrowed down the problem to: a) akonadi, that grows in size after time goes by (thus it seems to have a leak?) b) qtwebkit. Whenever the system behaves oddly then I found a zombi process of qtwebkit for mercurio, and about 10 instances of qt webkit running. c) when this happens (also without suspend to disc, it is sufficient to have
In data domenica 10 giugno 2018 08:09:48 CEST, hai scritto: the system idle for some time) you have the wlan that ceases to work. Oddly it says it is active and working but network does not seem to resolve addresses. I will try now the work around to define a 8.8.8.8 as obligatory dns in nw- manager. Maybe this can help with the last point. Currently I am not positively impressed with the quality of this release, I am honest. If asked i would advice to stay with 42.3 currently as long as possible. _________________________________________________________________ ________________________________________________________ Ihre E-Mail-Postfächer sicher & zentral an einem Ort. Jetzt wechseln und alte E-Mail-Adresse mitnehmen! https://www.eclipso.de -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
* stakanov <stakanov@eclipso.eu> [06-12-18 16:21]: [...]
I tried with akonadictl restart, or with akonadictl stop I have narrowed down the problem to: a) akonadi, that grows in size after time goes by (thus it seems to have a leak?)
move and rebuild the akonadi database. it may be a squirrelly entry.
b) qtwebkit. Whenever the system behaves oddly then I found a zombi process of qtwebkit for mercurio, and about 10 instances of qt webkit running.
don't know what mecurio is but, replace or disable it until other problems are solved. then run it with debug or trace.
c) when this happens (also without suspend to disc, it is sufficient to have the system idle for some time) you have the wlan that ceases to work. Oddly it says it is active and working but network does not seem to resolve addresses. I will try now the work around to define a 8.8.8.8 as obligatory dns in nw- manager. Maybe this can help with the last point. Currently I am not positively impressed with the quality of this release, I am honest. If asked i would advice to stay with 42.3 currently as long as possible.
what are your network settings? -- (paka)Patrick Shanahan Plainfield, Indiana, USA @ptilopteri http://en.opensuse.org openSUSE Community Member facebook/ptilopteri Registered Linux User #207535 @ http://linuxcounter.net Photos: http://wahoo.no-ip.org/piwigo paka @ IRCnet freenode -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
Op dinsdag 12 juni 2018 23:22:25 CEST schreef Patrick Shanahan:
* stakanov <stakanov@eclipso.eu> [06-12-18 16:21]: [...]
I tried with akonadictl restart, or with akonadictl stop I have narrowed down the problem to: a) akonadi, that grows in size after time goes by (thus it seems to have a leak?)
move and rebuild the akonadi database. it may be a squirrelly entry.
b) qtwebkit. Whenever the system behaves oddly then I found a zombi process of qtwebkit for mercurio, and about 10 instances of qt webkit running. don't know what mecurio is but, replace or disable it until other problems are solved. then run it with debug or trace.
c) when this happens (also without suspend to disc, it is sufficient to have the system idle for some time) you have the wlan that ceases to work. Oddly it says it is active and working but network does not seem to resolve addresses. I will try now the work around to define a 8.8.8.8 as obligatory dns in nw- manager. Maybe this can help with the last point. Currently I am not positively impressed with the quality of this release, I am honest. If asked i would advice to stay with 42.3 currently as long as possible.
what are your network settings? FWIW: I have moved from mariadb to postgresql about half a year ago and never regretted doing so one single day. It requires libQt5Sql5-postgresql to be installed and a change to akonadiserverrc, in my case QPSQL] Host=/tmp/akonadi-knurpht.RANDOM_STRING InitDbPath=/usr/bin/initdb Name=akonadi Options= ServerPath=/usr/bin/pg_ctl StartServer=true
Searches are fast, the entire handling of > 200.000 emails as well. -- Gertjan Lettink a.k.a. Knurpht openSUSE Board Member openSUSE Forums Team -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
In data martedì 12 giugno 2018 23:22:25 CEST, Patrick Shanahan ha scritto: Hallo Patrick
* stakanov <stakanov@eclipso.eu> [06-12-18 16:21]: [...]
I tried with akonadictl restart, or with akonadictl stop I have narrowed down the problem to: a) akonadi, that grows in size after time goes by (thus it seems to have a leak?)
move and rebuild the akonadi database. it may be a squirrelly entry.
b) qtwebkit. Whenever the system behaves oddly then I found a zombi process of qtwebkit for mercurio, and about 10 instances of qt webkit running. don't know what mecurio is but, replace or disable it until other problems are solved. then run it with debug or trace. Mercurio is my user id that runs all the pop mail tasks (that is physically separated from e.g. entropia where I do surf and mailing list. Rational is to avoid to a max having that user compromised as it manages e.g. online shopping tasks (payments) and private mail. I am ultra extremely compartimental when setting up my systems. Makes life also more ordered. The problem appears to be related with powersaving here too. Once the system is idle for a long time, something crashes or leaks. Sometimes the memory gets huge or... the temp goes up until emergency shutdown. In top in these cases you see often zombies of qt webkit. I found out also the following: when the ram is high, shutdown of kmail/ kontact does not make it better. But akonadictl stop makes it a bit better. When doing "exit kontact, then doing akonadictl and then (and here we go) you start kontact again, this will segfault silently. The start of kontact right hereafter will succeed instead and the memory will be normal again... If you can give me some hints about how to catch what is happening, would be great for a report because I can reproduce this quite easily (it is sufficient to let the system idle long enough.
c) when this happens (also without suspend to disc, it is sufficient to have the system idle for some time) you have the wlan that ceases to work. Oddly it says it is active and working but network does not seem to resolve addresses. I will try now the work around to define a 8.8.8.8 as obligatory dns in nw- manager. Maybe this can help with the last point. Currently I am not positively impressed with the quality of this release, I am honest. If asked i would advice to stay with 42.3 currently as long as possible.
what are your network settings? Well this bug is unrelated to kontact or, related in the sense that it is
Sorry in what sense? Move to a new user account? I found the filters being probably the major problem. Currently I substitute them one by one setting up a copy and deleting the old ones. Seems to work to some extend. triggered by suspend to disk and especially when the system is running but is idle for a long time (e.g. 2 - 3 hours. That is the point, they are very very standard. Today I had incoming updates to pam, will see if this makes a difference. It appears that there is a permission problem? together with a wake up problem and in the latter aspect especially it seems that the wlan cannot access the kwallet when waking up. I am still trying to understand, but it goes in that direction. So it could also be a race condition, that e.g. kwallet wakes up too late for wlan that tries already to reconnect. A part of this, firewall is very restrictive (literally nothing allowed but surfing and mail) and ipv6 is set to ignore. But ipv4 vs 6 should make no difference as when you start the network works well and connects (with that strange ritual that first it asks for password root, then it says "no secrets where given" and "deactivated wlan" then it asks again(!) the root password, then wallet comes up and then ..... it connects. But it does so also if you just press enter or give a wrong root password. Now I filed a bug time ago for 42.3 on this behalf and did take it to 15 only yesterday, describing the problem. Let us see if somebody finds the time to respond. If not: dns is set to be taken from the router (which has a double entry, if one would fail. Resolve.conf appears normal and does not complain. In journalctl I found something about complaining "empty window for wallet" so I think there is the problem. In the meanwhile, thank you for asking. _________________________________________________________________ ________________________________________________________ Ihre E-Mail-Postfächer sicher & zentral an einem Ort. Jetzt wechseln und alte E-Mail-Adresse mitnehmen! https://www.eclipso.de -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
participants (7)
-
David C. Rankin -
John Andersen -
Knurpht ＠ openSUSE -
Patrick Shanahan -
Per Jessen -
Richard Brown -
stakanov