I''ve setup a new mongrel cluster on my VPS, and am noticing a lot if 503 errors if I try to hit the server too many times in succession in a few seconds. I''ve read that this is usually caused by the cluster crashing, but I haven''t been seen the cluster crash when I do this. Does anyone have any other ideas what might cause this? I have my server configured how MediaTemple suggested (http://kb.mediatemple.net/article.php?id=279). It just seems like Apache isn''t able to always see the cluster. Thanks for your help. - Justin Williams Owner, Second Gear http://www.secondgearllc.com/
I get that when I start up apache and there isn''t a mongrel available to answer. I think in your case it may be that all your mongrels are busy, and it''s letting you know that. You might need to add more mongrel processes to see if it helps, but that may be an issue if you''re only running on a VPS with 64MB.... Mongrel w/rails starts at 32mb and grows for me, so it seems you could have 2 running before you get into performance issues... On 7/6/07, Justin Williams <carpeaqua at gmail.com> wrote:> I''ve setup a new mongrel cluster on my VPS, and am noticing a lot if > 503 errors if I try to hit the server too many times in succession in > a few seconds. I''ve read that this is usually caused by the cluster > crashing, but I haven''t been seen the cluster crash when I do this. > > Does anyone have any other ideas what might cause this? I have my > server configured how MediaTemple suggested > (http://kb.mediatemple.net/article.php?id=279). It just seems like > Apache isn''t able to always see the cluster. > > Thanks for your help. > > - > Justin Williams > Owner, Second Gear > http://www.secondgearllc.com/ > _______________________________________________ > Mongrel-users mailing list > Mongrel-users at rubyforge.org > http://rubyforge.org/mailman/listinfo/mongrel-users >
VPS you say? How much memory do you have allocated to you? Rails takes A LOT of memory. I''ve seen Mongrels running rails eating up 20 Mb EACH for *very* simple actions. I recall people on this list mentioning 50+ Mb when RMagick is involved. Make sure you''re not hitting the memory limit on your VPS, which triggers killing off the process, or parts thereof. =Will Green Joey Geiger wrote:> I get that when I start up apache and there isn''t a mongrel available to answer. > > I think in your case it may be that all your mongrels are busy, and > it''s letting you know that. You might need to add more mongrel > processes to see if it helps, but that may be an issue if you''re only > running on a VPS with 64MB.... Mongrel w/rails starts at 32mb and > grows for me, so it seems you could have 2 running before you get into > performance issues... > > On 7/6/07, Justin Williams <carpeaqua at gmail.com> wrote: >> I''ve setup a new mongrel cluster on my VPS, and am noticing a lot if >> 503 errors if I try to hit the server too many times in succession in >> a few seconds. I''ve read that this is usually caused by the cluster >> crashing, but I haven''t been seen the cluster crash when I do this. >> >> Does anyone have any other ideas what might cause this? I have my >> server configured how MediaTemple suggested >> (http://kb.mediatemple.net/article.php?id=279). It just seems like >> Apache isn''t able to always see the cluster. >> >> Thanks for your help. >> >> - >> Justin Williams >> Owner, Second Gear >> http://www.secondgearllc.com/ >> _______________________________________________ >> Mongrel-users mailing list >> Mongrel-users at rubyforge.org >> http://rubyforge.org/mailman/listinfo/mongrel-users >> > _______________________________________________ > Mongrel-users mailing list > Mongrel-users at rubyforge.org > http://rubyforge.org/mailman/listinfo/mongrel-users
I''ve got three Mongrels spawned, each hovering around 40MB each. I also just checked memory usage, and that doesn''t seem to be an issue at all. When I try to pound the server, it''s not killing off the processes, which is what is weird to me. Usually when I get the 503 error it is because mongrel isn''t spawned, but in this case all the processes are up and running fine, and the connection works about half the time. It''s just every few actions it will fail. Any other ideas? - j On 7/6/07, Will Green <will at hotgazpacho.com> wrote:> VPS you say? How much memory do you have allocated to you? Rails takes A > LOT of memory. I''ve seen Mongrels running rails eating up 20 Mb EACH for > *very* simple actions. I recall people on this list mentioning 50+ Mb > when RMagick is involved. > > Make sure you''re not hitting the memory limit on your VPS, which > triggers killing off the process, or parts thereof. > > => Will Green > > Joey Geiger wrote: > > I get that when I start up apache and there isn''t a mongrel available to answer. > > > > I think in your case it may be that all your mongrels are busy, and > > it''s letting you know that. You might need to add more mongrel > > processes to see if it helps, but that may be an issue if you''re only > > running on a VPS with 64MB.... Mongrel w/rails starts at 32mb and > > grows for me, so it seems you could have 2 running before you get into > > performance issues... > > > > On 7/6/07, Justin Williams <carpeaqua at gmail.com> wrote: > >> I''ve setup a new mongrel cluster on my VPS, and am noticing a lot if > >> 503 errors if I try to hit the server too many times in succession in > >> a few seconds. I''ve read that this is usually caused by the cluster > >> crashing, but I haven''t been seen the cluster crash when I do this. > >> > >> Does anyone have any other ideas what might cause this? I have my > >> server configured how MediaTemple suggested > >> (http://kb.mediatemple.net/article.php?id=279). It just seems like > >> Apache isn''t able to always see the cluster. > >> > >> Thanks for your help. > >> > >> - > >> Justin Williams > >> Owner, Second Gear > >> http://www.secondgearllc.com/ > >> _______________________________________________ > >> Mongrel-users mailing list > >> Mongrel-users at rubyforge.org > >> http://rubyforge.org/mailman/listinfo/mongrel-users > >> > > _______________________________________________ > > Mongrel-users mailing list > > Mongrel-users at rubyforge.org > > http://rubyforge.org/mailman/listinfo/mongrel-users > _______________________________________________ > Mongrel-users mailing list > Mongrel-users at rubyforge.org > http://rubyforge.org/mailman/listinfo/mongrel-users >-- - Justin Williams justin at carpeaqua.com work: http://www.secondgearllc.com/ play: http://www.carpeaqua.com
I am not an Apache expert, but I have had much success with the method preached on the mongrel site: http://mongrel.rubyforge.org/docs/apache.html for using mod proxy balancer. You can mess with the timeout settings very easily there, so apache will wait a bit longer before reporting back a 503. I looked at the instructions on mediatemple, and they seemed a bit inadequate. Do you have Apache 2.2.3 on your VPS? On 7/6/07, Justin Williams <carpeaqua at gmail.com> wrote> > I''ve got three Mongrels spawned, each hovering around 40MB each. I > also just checked memory usage, and that doesn''t seem to be an issue > at all. > > When I try to pound the server, it''s not killing off the processes, > which is what is weird to me. Usually when I get the 503 error it > is because mongrel isn''t spawned, but in this case all the processes > are up and running fine, and the connection works about half the time. > It''s just every few actions it will fail. > > Any other ideas? > > - j > > On 7/6/07, Will Green <will at hotgazpacho.com> wrote: > > VPS you say? How much memory do you have allocated to you? Rails takes A > > LOT of memory. I''ve seen Mongrels running rails eating up 20 Mb EACH for > > *very* simple actions. I recall people on this list mentioning 50+ Mb > > when RMagick is involved. > > > > Make sure you''re not hitting the memory limit on your VPS, which > > triggers killing off the process, or parts thereof. > > > > => > Will Green > > > > Joey Geiger wrote: > > > I get that when I start up apache and there isn''t a mongrel available > to answer. > > > > > > I think in your case it may be that all your mongrels are busy, and > > > it''s letting you know that. You might need to add more mongrel > > > processes to see if it helps, but that may be an issue if you''re only > > > running on a VPS with 64MB.... Mongrel w/rails starts at 32mb and > > > grows for me, so it seems you could have 2 running before you get into > > > performance issues... > > > > > > On 7/6/07, Justin Williams <carpeaqua at gmail.com> wrote: > > >> I''ve setup a new mongrel cluster on my VPS, and am noticing a lot if > > >> 503 errors if I try to hit the server too many times in succession in > > >> a few seconds. I''ve read that this is usually caused by the cluster > > >> crashing, but I haven''t been seen the cluster crash when I do this. > > >> > > >> Does anyone have any other ideas what might cause this? I have my > > >> server configured how MediaTemple suggested > > >> (http://kb.mediatemple.net/article.php?id=279). It just seems like > > >> Apache isn''t able to always see the cluster. > > >> > > >> Thanks for your help. > > >> > > >> - > > >> Justin Williams > > >> Owner, Second Gear > > >> http://www.secondgearllc.com/ > > >> _______________________________________________ > > >> Mongrel-users mailing list > > >> Mongrel-users at rubyforge.org > > >> http://rubyforge.org/mailman/listinfo/mongrel-users > > >> > > > _______________________________________________ > > > Mongrel-users mailing list > > > Mongrel-users at rubyforge.org > > > http://rubyforge.org/mailman/listinfo/mongrel-users > > _______________________________________________ > > Mongrel-users mailing list > > Mongrel-users at rubyforge.org > > http://rubyforge.org/mailman/listinfo/mongrel-users > > > > > -- > - > Justin Williams > justin at carpeaqua.com > work: http://www.secondgearllc.com/ > play: http://www.carpeaqua.com > _______________________________________________ > Mongrel-users mailing list > Mongrel-users at rubyforge.org > http://rubyforge.org/mailman/listinfo/mongrel-users >-- -Matt McKnight- www.lmnsolutions.com -------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20070706/86b5e8ca/attachment.html
Switching over to mod_proxy_balancer seems to help things out. Thanks for the help! - j On 7/6/07, Matthew McKnight <matt.mcknight at gmail.com> wrote:> I am not an Apache expert, but I have had much success with the method > preached on the mongrel site: > http://mongrel.rubyforge.org/docs/apache.html for using mod > proxy balancer. You can mess with the timeout settings very easily there, > so apache will wait a bit longer before reporting back a 503. I looked at > the instructions on mediatemple, and they seemed a bit inadequate. Do you > have Apache 2.2.3 on your VPS? > > > > On 7/6/07, Justin Williams <carpeaqua at gmail.com> wrote > > I''ve got three Mongrels spawned, each hovering around 40MB each. I > > also just checked memory usage, and that doesn''t seem to be an issue > > at all. > > > > When I try to pound the server, it''s not killing off the processes, > > which is what is weird to me. Usually when I get the 503 error it > > is because mongrel isn''t spawned, but in this case all the processes > > are up and running fine, and the connection works about half the time. > > It''s just every few actions it will fail. > > > > Any other ideas? > > > > - j > > > > On 7/6/07, Will Green < will at hotgazpacho.com> wrote: > > > VPS you say? How much memory do you have allocated to you? Rails takes A > > > LOT of memory. I''ve seen Mongrels running rails eating up 20 Mb EACH for > > > *very* simple actions. I recall people on this list mentioning 50+ Mb > > > when RMagick is involved. > > > > > > Make sure you''re not hitting the memory limit on your VPS, which > > > triggers killing off the process, or parts thereof. > > > > > > => > > Will Green > > > > > > Joey Geiger wrote: > > > > I get that when I start up apache and there isn''t a mongrel available > to answer. > > > > > > > > I think in your case it may be that all your mongrels are busy, and > > > > it''s letting you know that. You might need to add more mongrel > > > > processes to see if it helps, but that may be an issue if you''re only > > > > running on a VPS with 64MB.... Mongrel w/rails starts at 32mb and > > > > grows for me, so it seems you could have 2 running before you get into > > > > performance issues... > > > > > > > > On 7/6/07, Justin Williams <carpeaqua at gmail.com > wrote: > > > >> I''ve setup a new mongrel cluster on my VPS, and am noticing a lot if > > > >> 503 errors if I try to hit the server too many times in succession in > > > >> a few seconds. I''ve read that this is usually caused by the cluster > > > >> crashing, but I haven''t been seen the cluster crash when I do this. > > > >> > > > >> Does anyone have any other ideas what might cause this? I have my > > > >> server configured how MediaTemple suggested > > > >> (http://kb.mediatemple.net/article.php?id=279). It > just seems like > > > >> Apache isn''t able to always see the cluster. > > > >> > > > >> Thanks for your help. > > > >> > > > >> - > > > >> Justin Williams > > > >> Owner, Second Gear > > > >> http://www.secondgearllc.com/ > > > >> _______________________________________________ > > > >> Mongrel-users mailing list > > > >> Mongrel-users at rubyforge.org > > > >> http://rubyforge.org/mailman/listinfo/mongrel-users > > > >> > > > > _______________________________________________ > > > > Mongrel-users mailing list > > > > Mongrel-users at rubyforge.org > > > > http://rubyforge.org/mailman/listinfo/mongrel-users > > > _______________________________________________ > > > Mongrel-users mailing list > > > Mongrel-users at rubyforge.org > > > http://rubyforge.org/mailman/listinfo/mongrel-users > > > > > > > > > -- > > - > > Justin Williams > > justin at carpeaqua.com > > work: http://www.secondgearllc.com/ > > play: http://www.carpeaqua.com > > _______________________________________________ > > Mongrel-users mailing list > > Mongrel-users at rubyforge.org > > http://rubyforge.org/mailman/listinfo/mongrel-users > > > > > > -- > -Matt McKnight- > www.lmnsolutions.com > _______________________________________________ > Mongrel-users mailing list > Mongrel-users at rubyforge.org > http://rubyforge.org/mailman/listinfo/mongrel-users >-- - Justin Williams justin at carpeaqua.com work: http://www.secondgearllc.com/ play: http://www.carpeaqua.com