On 04/11/17 09:55 PM, Nathan Stratton Treadway wrote: > On Mon, Oct 16, 2017 at 09:20:16 -0400, Jean-Louis Martineau wrote: > > Amanda 3.5 can do everything you want only by running the amdump > command. > > > > Using an holding disk: > > > > * You configure two storages > > * All dumps go to the holding disk > > * All dumps are copied to each storages, not necessarily at the same > > time or in the same run. > > * The dumps stay in holding until they are copied to both storages > > * You can tell amanda that everything must go to both storage or only > > some dle full/incr > > > > Without holding disk > > > > * You configure one storage telling it to automatically vault its > > dumps to the vault-storage > > * You configure one vault-storage > > * All dumps are dumped to the storage then vaulted to the vault-storage > > * The vault operation can be done later in the same run or in a > > different run. > > * All dumps must go to the storage, but you can vault only some dles > > full/incr > > Jean-Louis, are the following correct summararies of how one would > actually implement the two scenarios you describe above? > > 1) the "using a holding disk" scenario would involve: > * listing both storages in the "storage" global parameter > * no "vault-storage" line > * no "vault" parameter in either storage definition > * the "you can tell amanda that everything...." part would be > accomplished using the "dump-selection" parameter in each storage > definition YES > > 2) the "without holding disk" scenario would involve: > * no global "storage" parameter (or one listing a single storage) It must list the main storage. > * the "vault" parameter would be found in the storage definition for > that main storage > * if you want to do the vaulting as part of the amdump run, the > global "vault-storage" parameter would list the (single) vault storage > * if you want to do the vaulting later, you would not have > "vault-storage" defined (and would kick off the vaulting operation with > amdump CONF BADHOST -ovault-storage="VAULT" > ) > * the "you can vault only some dles full/incr" part would be > accomplished using the "dump-selection" parameter in the vault > storage definition > * if a holding disk is defined, dumps stay there only until they are > written to the main storage, after which they are deleted from the > holding disk (and thus the vaulting step pulls dumps back off the > storage to which they were previously written) YES > > > > If I'm using vtapes (on internal drives) for my main storage, is there > any downside to using the second scenario instead of the first? This scenario can be better because it reduce the number of read/write to the server disk. But it can also reduce the parallelism, so you should increase taper-parallel-write and also configure amanda so than it use more than one tape. Setting 'max-dle-by-volume 1' and adding a lot of slots can achieve it. > > (I am already using a holding disk for the main dump, but it seems like > using the second scenario would mean that if the vault storage isn't > available during the run, I won't end up filling the holding disk with > dumps that have already been written successfully to the main > storage....) > > > > Finally, a couple question about your statement "All dumps are copied to > each storages, not necessarily at the same time or in the same run." (in > the use-holding-disk scenario): > > * am I correct that amdump would always try to use both storages, and > thus situation in which they would be written at different times is > when an acceptable volume is not found when amdump starts? They are never written at the same time! Each storage have different setting and speed. They are written in the same run if the configuration allow and acceptable volumes are found. > > (Does having one of the two storages with no volume found drop the > entire run into degraded mode?) Good question, I think no (If I remember correctly) > > * Once that has happened, one would make the missing volume available > and then use the "amflush" command to kick off the pending writes, > right? (rather than "amdump CONF BADHOST" as used in the > no-holding-disk scenario) yes
Jean-Louis This message is the property of CARBONITE, INC. and may contain confidential or privileged information. If this message has been delivered to you by mistake, then do not copy or deliver this message to anyone. Instead, destroy it and notify me by reply e-mail
